The QPU scheduler allows to move certain TMU instructions around and
since we enabled pipelining, we need to protect against the case where
doing this might break a TMU sequence. For example, this test:
dEQP-VK.rasterization.line_continuity.line-strip
Was generating this VIR:
mov tmud, t187
mov.pushz null, t176
mov.ifa tmua, t9
nop null; wrtmuc (img[0].p0 | 0x0)
mov tmut, t185
mov tmud, t180
mov.ifa tmusf, t183
nop null; thrsw
where we have a general TMU access (tmud,tmua) followed by an image
access (wrtmuc, tmut, tmud, tmusf), which the QPU scheduler was turning
into:
nop ; nop ; ldunifrf.rf22 (0xffffff00 / -nan)
nop ; nop ; wrtmuc (img[0].p0 | 0x0)
nop ; nop ; ldtmu.r2
add r0, r2, 1 ; nop ; ldtmu.r3
nop ; nop ; ldtmu.r4
nop ; mov tmud, r0
nop ; mov.ifa tmua, rf15
nop ; mov tmut, r4 ; thrsw
nop ; mov tmud, rf22
nop ; mov.ifa tmusf, r3
where it allowed the wrtmuc to move up and before the general TMU access,
leading to an incorrect TMU sequence.
Fix this by flagging TMUA writes (which are the sequence terminators for
general TMU accessess) as writing new TMU configuration, like we do for all
other TMU sequence terminators for textures and images.
Fixes: 197090a3fc ('broadcom/compiler: implement pipelining for general TMU operations')
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8954>
this just flushes all the pending batches, which is enough since we
do our batch work synchronously and everything is awful
will be reworked when this is no longer the case
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8948>
viewport can be transformed at draw time based on clip_halfz, so we actually
want to do the conversion from gallium->vulkan at that time
as a nice bonus, we get to delete a struct member
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8946>
If the device isn't lost, the application might want to re-submit.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8931>
Fixes the documentation preview link on the MR Overview tab. Found by
comparing with the same thing in weston.
(It still appears only after all jobs in the pipeline have at least
started once though)
Fixes: ae8f9584f4 "CI: always expose docs artifacts"
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8927>
These offsets were wrong and didn't match the old behaviour.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: e8220e106b ("aco: optimize AC_FETCH_FORMAT_SNORM alpha adjust")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8935>
On GFX10+, the ESGS ring is allocated in LDS. This saves one BO
when geometry shaders are used, and possibly two BOs if nothing
else (tess, scratch, etc) is used.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8926>
It's already freed in radv_destroy_fence().
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8938>
The GL frontend can lower this weird GL feature away for us. This should
fix redeclaration of the gl_Color/SecondaryColor as centroid, since that
case had been missed in the !flat special case here.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8601>
We had an unnecessary case in our uniforms upload switch statement, since
we no longer advertise the cap.
Fixes: 8ad931808e ("v3d: do not report alpha-test as supported")
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8601>
We stopped reporting the alpha test screen cap, and stopped using the
value in the key, so now shrink the key. This gets another switch case
out of the hot uniforms upload path.
Fixes: 1404b8b1e5 ("vc4: do not report alpha-test as supported")
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8601>
Since we always precompile now, we can remove some checks for presence of
the first variant or for it being the wrong key.
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8601>
Every driver wants the chance to see the default variant of the shader in
the hopes of avoiding draw-time shader compiles, even if they have some
NOS compiled into the shader by st/mesa. Before, every driver had to
avoid st/mesa lowering options to get this behavior, even if they would
like to have the shared lowering code, and some drivers were just missing
the behavior because nobody had debugged that it was missing yet.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8601>
The version overrides leaked into the virglrenderer server, breaking
compiling of all sorts of shaders. We already get GLES3.1 when running
virgl on llvmpipe GL without doing any overrides.
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8854>
if compute or gfx use a resource which is already in use by the other pipeline,
we need to ensure tha we flush that pipeline now in order to have commands execute
in the expected order
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
this is almost identical, the difference being that we need to explicitly
use the compute batch/program in a few places
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
this is a bit different since the regular functions cycle the gfx batch
array, but the compute batch isn't in an array
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
this is a massive block of code that we need to reuse for compute shaders
because we absolutely don't want to be copying it around
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
these are similar to the gfx variants, but mostly smaller
some things are implemented as placeholders for now to make the code
similar to gfx programs/pipelines and thus easier to read/review
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
by moving this into its own function, we can reuse it for creating a compute
batch, which will not be part of the gfx batch array
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
this lets us reuse the function for gfx and compute programs without much
churn, since in C we can use a pointer to the first member of a struct and
a pointer to the struct interchangeably
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8781>
