anv: Avoid sampling some MCS surfaces with clear

Supposedly avoids GPU hangs in BF4. See HSD 1707282275 and 14013111325.

v2. Fix bug in WA implementation. (Sagar)

Cc: mesa-stable
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8755>

anv: Add clear_supported to anv_layout_to_aux_state

This will be used for an MCS workaround.

Cc: mesa-stable
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8755>

radv: call nir_copy_prop() after load-store vectorization

The load-store vectorizer can create a large amount
of unnecessary nir_op_vec and nir_op_mov instructions.
This prevents nir_opt_move from stalling to much and
potentially also helps other passes.

Closes: #4778
Fixes: 1958381c9ae15dc252bcab8612f39fdca45d4843 ('radv: Reorder some NIR optimizations in preparation for the I/O changes.')
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10804>

aco/ra: prevent underflow register for p_create_vector operands

It could happen that we tested negative out-of-range
registers for p_create_vector operands resulting in a crash.

Fixes: 8962510e38fbaff792f60bc17d46507bb77401ac ('aco/ra: Conservatively refactor get_reg_specified to use PhysRegInterval')
Closes: #4697
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10799>

ci/freedreno: Skip refract on a306 now that it hangchecks sometimes.

Not every MR, but several per day since I landed the apitrace switch which
increased trace replay resolution to 1080p.  Hopefully some day we can
tune the hangcheck to be less aggressive.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10793>

nir: Rename nir_is_per_vertex_io to nir_is_arrayed_io

VS outputs are "per vertex" but not the kind of I/O we want to match
with this helper.  Change to a name that covers the "arrayness"
required by the type.

Name inspired by the GLSL spec definition of arrayed I/O.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10493>

compiler/nir: check whether var is an input in lower_fragcoord_wtrans

Otherwise the lowering pass might try to lower any other load from
a deref if its data.location value happens to be zero.

Fixes: 418c4c0d7d48a42f475df1ffb93b3a33763e7a4a
  compiler/nir: extend lower_fragcoord_wtrans to support VARYING_SLOT_POS

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10577>

zink: emit some barriers out of renderpass where possible

when emitting barriers, we don't need to end the renderpass in some cases

this handles the case of emitting barriers for new resources which have
never before been accessed and thus do not require synchronization at a
specific point in the batch

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10700>

zink: add flags for determining whether to update framebuffer and renderpass

we can optimize restarting renderpasses here a little by only performing
lookups if states have actually changed

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10761>

zink: add a flag for tracking/validating renderpass clears

we've been tracking which attachments are deferring clears but not
specifically the ones that are going to be cleared in a renderpass

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10761>

zink: split out base renderpass begin into separate function

let fb setup and barriers and whatever be separate

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10761>

util: tune signatures of generated enum operators

This allows use of these operators in a constexpr context and silences
a clang -Wunused-function warning.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10784>

util: add support for defining bitwise operators on strongly typed enums

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10784>

intel: Add 2 ADL-S pci-ids

Cc: mesa-stable
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10781>

driconf: set vk_x11_strict_image_count for Metro: Exodus

Otherwise, the game crashes on startup under xwayland.

Closes: #4650
Cc: mesa-stable
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10750>

util/queue: don't require a fence when adding a job

sometimes we just want to fire jobs off into the void and don't care
if or when they finish

Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10719>

panfrost: Skip blit shader labelling if the buffer has no space

Fixes stack corruption in dEQP-GLES31.functional.draw_buffers_indexed.
random.max_implementation_draw_buffers.10

Fixes: 8ba2f9f6985 ("panfrost: Create a blitter library to replace the existing preload helpers")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10792>

anv: handle spirv parsing failure

v2: don't leak spec_entries

v3: Also switch to VK_ERROR_UNKNOWN when parsing fails

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10790>

intel/dev: Add device info for ADL GT2

Cc: mesa-stable
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9465>

intel/isl: Add Wa_22011186057 to disable CCS on ADL GT2 A0

Cc: mesa-stable
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9465>

CI: Disable all Panfrost/AMD/Iris automatic jobs

The power in Cambridge is unstable; disable these jobs until it's back
clear.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10797>

zink: immediately return false when getting query result if it's not gonna happen

this is a small optimization for the no-wait case when unflushed usage
exists since it's impossible for a qbo update to happen instantly

no functionality will be fixed by this, it's just a very minor optimization

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10796>

lavapipe: fix fencing when submitting multiple cmdbufs

a fence applies to all the submitted cmdbufs, so it's necessary to do
the flush which creates the user fence after all the cmdbufs have been
processed in order to avoid creating a fence that only applies to the
first cmdbuf

Fixes: b38879f8c5f ("vallium: initial import of the vulkan frontend")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10795>

zink: fix DrawParameters shader cap usage

this is for vertex shaders and covers more than just baseinstance

Fixes: dcb9e4ddb4b ("zink: emit cap early")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10788>

zink: stop overwriting buffer map pointers for stream uploader

this breaks the driver!

the uploader always maps its own pointer, so modifying that at any
point just explodes things later

Fixes: d179c5d28e6 ("zink: implement threaded context")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10787>

intel/compiler: Clarify why VUE is recomputed by FS

FS will get the last geometry VUE, but it still needs to recompute in
case the number of position slots assigned by geometry is larger than
one -- this happens when Primitive Replication is used.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10653>

ci/turnip: drop fail annotation for image.extend_operands_spirv1p4.*

They were fixed in
ed20e69b "vtn: Handle ZeroExtend/SignExtend image operands"

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10783>

ci/turnip: drop fail annotation for float_control tests

These tests are NotSupported and therefore cannot fail.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10783>

anv: fix perf query pass with command buffer batching

We've only considered the perf query pool change previously. But we
also need to pay attention to the pass index.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 0a7224f3ff7542 ("anv: group as many command buffers into a single execbuf")
Reviewed-by: Felix DeGrood <felix.j.degrood@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10301>

intel/mi_builder: fix resolve call

Giving NULL for anv_combine_address() triggers an assert in that
function.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 8525ebe6e375 ("intel/mi_builder: Return an address from __gen_get_batch_address")
Reviewed-by: Felix DeGrood <felix.j.degrood@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10301>

i915g: add HW atomic counters as unsupported

Closes: #4772
Fixes: 2a06423c009 ("gallium: add CAPs to support HW atomic counters. (v3)")
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10776>

aco/scheduler: Verify register demand invariants in debug mode

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10644>

aco/scheduler: Fix register demand computation for upwards moves

The initial value needs to be taken from the instruction that is being
moved over, not the one to be moved.

Additionally the parameter of this function was removed because it was
misleading. Setting it to any value other than source_idx would cause
register_demand to be initialized incorrectly. (Instead, the maximum
demand among the covered instructions would need to be determined.)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10644>

aco/scheduler: Fix register demand computation for downwards moves

Previously, changes in total_demand_clause were not always propagated to
total_demand. For instance, clause moves do not change the local register
demand at the end of a clause, yet they may still affect the total maximum.

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 8235bc64112 ("aco: try to group together VMEM loads of the same resource")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4533
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10644>

aco: relax validation rules for p_reduce dst RegType

By exposing a subgroupSize of 64, reductions with
cluster_size 32 in wave32 might be considered divergent,
and thus, result in a VGPR.

Fixes: dEQP-VK.subgroups.clustered.graphics.subgroupclustered* with wave32

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10769>

isl: require hiz for depth surface in isl_surf_get_ccs_surf

Fixes: 752eefdb ("intel/isl: Refactor isl_surf_get_ccs_surf")
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10768>

wgl: Remove opengl32.mingw.def.

MinGW DEF parsing is even more broken than before, and 32-bits import
libs are broken regardless one uses opengl32.mingw.def or opengl32.def.

This change removes opengl32.mingw.def and addresses the issue differently:

- link opengl32.dll with --enable-stdcall-fixup

- use the systems opengl32 import lib (libopengl32.a/opengl32.lib)
  instead of our own

This change also gets test_wgl built with MinGW (even if it's never tested),
which I used to verify this; and to not link against internal libraries.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
v2: Revert back to shared_library.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10767>

zink: always do maybe_flush after draw/compute

this is a bit more sensible since a forced flush is going to require
refs and layouts be reapplied

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10760>

zink: check for a work_count-based stall in zink_maybe_flush_or_stall()

put all the maybes in one place

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10760>

zink: flush every 100k draws/computes

this ensures more consistent throughput in e.g., drawoverhead

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10760>

glsl: add missing support for explicit components in interface blocks

From the ARB_enhanced_layouts spec:

   "As with input layout qualifiers, all shaders except compute shaders
   allow *location* layout qualifiers on output variable declarations,
   output block declarations, and output block member declarations.  Of
   these, variables and block members (but not blocks) additionally
   allow the *component* layout qualifier."

We previously had compile tests in piglit to make sure this was not a
compile error but no execution tests.

Fixes: d99a040bbf2c ("i965: enable ARB_enhanced_layouts for gen8+")

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10763>

glsl: create validate_component_layout_for_type() helper

This will be used in the following patch.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10763>

Update release notes with mention that x87 is no longer used on x86

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868>

Switch u_format_test to passed on i386

Switching from x87 to SSE resolves the rounding behaviour in this test.

It was some of the dxt formats failing and now they are not.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868>

Default enable SSE2 on mesa builds.

With the idea of branching classic device support in to its own tree now would be a good time to also raise the minimum
requirements to something that is more "modern" on x86.
SSE2 was introduced in 2000(!) by default let's make it the minimum spec now
All the old hardware that is moving to the maintenance branch will finally be out of the way.

For the 64-bit side of the discussion there isn't much changed.
* GCC already enables -msse and -msse2 by default
* Same with clang
* fpmath=sse might remove some extraneous x87 usage
** Clang implies fpmath=sse ALWAYS

For the 32-bit side of things is where the exciting details change
* GCC by default doesn't enable sse1 or sse2
** Does all `float`, `double`, and `long double` math with x87
** -msse2 enables sse2 and sse1, gcc still uses x87 even with those enabled
** -mfpmath=sse moves away from using x87 and instead uses sse1 and sse2
* Clang already default enables sse1/sse2 which then turns on their implied fpmath=sse

What does this mean for users?
On Linux raises the default minimum processor spec to SSE2 supporting CPUs
* Intel requirements raise from P5 (1993) to Netburst (2000)
* AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003)
* Via requirements raise from C3(2001) to C7 (2005)

What does it mean for package maintainers?
For x86-64 distributions that have i386/i686 multilib, then nothing changes. You're already on a platform guaranteed to support SSE2.
For i386/i686 distributions they will need to weigh their min spec against this. Not sure how many still support classic processors.

Who is left out in the cold?
* Intel Quark (2013)
** Embedded board, doesn't have a GPU, Technically has 1x PCIe 2.0 lane that someone could plug a GPU in to
* Some older transmeta CPUs, but they had a followup that also had SSE2.
** Anyone hacking on these with a modern GPU? I'm guessing they know how to turn this option off

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868>

nv50/ir: Add ConstantFolding constructor.

Fix defect reported by Coverity Scan.

Uninitialized scalar field (UNINIT_CTOR)
member_not_init_in_gen_ctor: The compiler-generated constructor for this
class does not initialize foldCount.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10713>

nvc0/ir: Initialize CodeEmitterGK110 member progType in constructor.

Fix defect reported by Coverity Scan.

Uninitialized scalar field (UNINIT_CTOR)
uninit_member: Non-static class member progType is not initialized in
this constructor nor in any functions that it calls.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10715>

venus: get rid of #ifdef's in vn_CreateImage

No real change after compiler optimizations.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10779>

intel/fs: use the final destination type for regioning restrictions

This is most likely a rebase mistake :(

Fixes: f3e5cd813ae6a8 ("intel/fs: Handle regioning restrictions of split FP/DP pipelines.")
Ref: aa53665fda6348 ("intel/fs/copy_prop: check stride constraints with actual final type")
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10764>

panfrost: Add GPU IDs for G52 1-Core-2EE (RK3568/RK3566)

Rockchip SoCs RK3566 and RK3568 have a Gondul with
one shader core and two execution engines, with product ID 0x7402.

Quoting the datasheet:

Mali-G52 1-Core-2EE
* Support 1600Mpix/s fill rate when 800MHz clock frequency
* Support 38.4GLOPs when 800MHz clock frequency

To distinguish it from other variants of G52, we agreed
to call this "G52L", L is for Little.

Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10771>

pan/mdg: Use smaller LD_UNIFORM instructions

If we only read 8 bytes from a UBO, we need to use LD_UNIFORM.64 as
opposed to LD_UNIFORM.128. In addition to probably being faster, this
fixes a buffer overrun manifesting as MMU faults with source ID
0x500/0x600/0x700, visible in WebGL Aquarium.

This is essentially the same patch as 616394cf31c ("pan/mdg: Use
appropriate sizes for global loads/stores"), only this is for UBOs where
that was for SSBOs.

Before enabling PACKED_UNIFORMS, this bug was not visible since we could
guarantee the UBO size was a multiple of 16. We no longer have that
invariant, and in rare cases the last 8 bytes of the last 16-byte slot
of a mapped uniform buffer would overrun the BO and trigger a fault,
even if the result is unused.

Fixes: 24d7c413fe7 ("panfrost: Enable packed uniforms.")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10772>

panfrost: Don't upload empty push uniform table

Instead zero out the pointer in this case.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10772>

panfrost: Clean up cases for emit_fbd

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10774>

panfrost: Remove spurious assignment

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10774>

panfrost: Don't translate compare funcs

Matches Gallium.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10774>

panfrost: Hide CAP_INT16 behind is_deqp

The GLSL precision lowering is still buggy (see !10729), no other driver
enables all the CAPs yet. I don't know enough GLSL IR to be the one to
fix this. In the mean time, this is a hotfix to expose the same set of
CAPs that radeonsi does.

By keeping it with is_deqp, we still get CI coverage of int16.

Closes: #4759
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10777>

freedreno/ir3: Don't force RTNE if rounding mode is undefined

Forcing round-to-nearest-even results in loss of opportunities for
conversion folding, causing a regression in gfxbench gl_alu2.

Fixes: de195671bdb ("ir3: nir_op_f2f16 should round to even")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10773>

frontends/va: fix multi planes for external memeory type

For the multi planes case, only the first plane is required with the
template buffer formats, and shouldn't fail for other planes.

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10751>

gallium/u_threaded: Add call logging

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10753>

gallium/u_threaded: Add to_call() helper

Add helper for casting the call, which when asserts are enabled will
sanity check the call size to detect corruption.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10753>

ci: Uprev piglit to 3351e8952 ("max-texture-size: report merged results")

This pulls in a fix for the max-texture-size test using piglit-runner,
among other things.

Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10749>

gallium/u_threaded: fix 32-bit breakage due to incorrect pointer arithmetic

Fixes: 1233c90ab4a - gallium/u_threaded: rewrite slot layout to reduce wasted space
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4755
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4758
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10746>

aco: fix additional register requirements for spilling

It could happen that VGPR spilling without SGPR spilling
calculated a negative spills_to_vgpr number and then
increasing the VGPR target demand above the limit.

Cc: mesa-stable
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10756>

radv: Use new NIR lowering of NGG GS when ACO is used.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

ac: Add NIR lowering for NGG GS.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

radv: Use new NGG NIR lowering for VS/TES when ACO is used.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

ac: Add new NIR pass to lower NGG VS/TES.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

aco: Implement new NGG specific NIR intrinsics.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

radv: Gather NGG info sooner.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

radv: Fill shader info earlier.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

aco: Split ngg_emit_sendmsg_gs_alloc_req from the wave0 check.

This allows us to emit the gs_alloc_req independently of the
wave ID check, which is what the NIR lowering will need.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

aco: Fixup the NIR metadata after sanitize_cf_list.

sanitize_cf_list can in fact invalidate the dominance metadata,
which we need to use eg. nir_unsigned_upper_bound.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

aco: Allow workgroup barrier and shared scope for NGG shaders.

NGG already needs to use workgroup barriers, but this commit allows
them to come from NIR as opposed to just emitting it in ACO
instruction selection.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

nir: Add AMD specific intrinsics for merged shaders and NGG.

These intrinsics represent what the hardware can actually do.
Lowering our shaders to use these intrinsics will allow us to
deal with mapping the classic VS, TES, GS (and the future MS)
stages to the hardware capabilities using NIR, which makes our
backend compilers simpler.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

nir: Allow load_primitive_id in VS in nir_divergence_analysis.

The lowered NIR code of NGG VS shaders uses this intrinsic
when the VS has to export the primitive ID.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

nir: Support upper bound of unsigned bit size conversions.

These allow us to generate slightly better code in some cases,
eg. multiplications in ACO.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

nir: Support upper bound of subgroup_id/num_subgroups for non-compute.

These intrinsics will be used when lowering NGG shaders, including
currently supported stages like VS, TES, GS and also by mesh shaders
in the future.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10740>

panfrost: Implement framebuffer invalidation

The strategy is lifted from Freedreno. The trick is to remove buffers
from the resolve set, rather than add buffers to a discard set (as you
would naively try). The latter is wrong -- draws after the
glInvalidateFramebuffer() still need to be respected.

glmark2 -btexture on-screen with sway on Mali T860 from 1393fps to
1998fps

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Closes: #2407
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6295>

panfrost: Set discard based on the resolve set

This is a bit simpler and will allow resolve to be disabled
independently.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6295>

panfrost: Track buffers needing resolve

This is a subset of the buffers cleared or drawed.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6295>

nir: handle float atomics in nir_lower_memory_model

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Fixes: 2adb337256f ("nir,radv/aco: add and use pass to lower make available/visible barriers")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10766>

nir: handle float atomics in nir_gather_info

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10766>

translate: assert that nr_elements is in range

This just ensures that we don't end up reading/writing outside of the
space reserved in the key. This would have made it easier to to track
down the issue in the previous commit faster.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10730>

translate: reserve more vertex-shader outputs

It's not just position that's special here, we also need space for
other special vertex-shader outputs.

This fixes a crash with Zink on Lavapipe, because Zink always emits
psize to avoid some shader-keys. While this might not be ideal, it's not
illegal. So we need to handle it.

But it seems we're missing more, and this list is taken from
compute_vertex_info() in lp_state_derived.c.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10730>

lavapipe: correct reported number of UBOs

When emitting the state we add one to the UBO index, so we have to
subtract one from the number of UBOs that we support. This is still
fine, because Vulkan only requires 14 UBOs to be supported, and LLVMpipe
supports 16.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10731>

panfrost: Flag indirect draw/dispatch shaders as internal

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10417>

panfrost: Expose the DRAW_PARAMETERS cap on Bifrost

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10417>

pan/bi: Add support for gl_DrawID

We simply pass the drawid as a sysval.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10417>

pan/bi: Add support for gl_{BaseVertex,BaseInstance}

Extend the VERTEX_INSTANCE_OFFSETS sysval to pass
BaseVertex/BaseInstance information to the shader.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10417>

pan/bi: Add first_vertex to vertex ID

Apparently r61 is zero based.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Tested-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10417>

panfrost: Split the indexed and !indexed indirect draw info structs

Some fields are overlapping, so let's just split those structs to
avoid confusion.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10417>

iris/ci: enable intel devices automatically in MR pipelines

After a period of stabilization it is now time to move them
to be part of the MR pipelines. We will still watch the
behavior and look for any issue that arises.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10550>

iris/ci: disable failing gimark test for now

In order to enable the intel devices to automatically be part of
Marge pipelines we need it to be fully passing all tests. We will
remove the gimark tests temporaly to achieve that.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10550>

nv50/ir: Initialize GCRA::RIG_Node members.

Fix defects reported by Coverity Scan.

uninit_member: Non-static class member degree is not initialized in this constructor nor in any functions that it calls.
uninit_member: Non-static class member degreeLimit is not initialized in this constructor nor in any functions that it calls.
uninit_member: Non-static class member maxReg is not initialized in this constructor nor in any functions that it calls.
uninit_member: Non-static class member f is not initialized in this constructor nor in any functions that it calls.
uninit_member: Non-static class member reg is not initialized in this constructor nor in any functions that it calls.
uninit_member: Non-static class member weight is not initialized in this constructor nor in any functions that it calls.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10714>

turnip: Implement VK_KHR_timeline_semaphore.

Implements non-shareable timelines using legacy syncobjs,
inspired by anv/radv implementation.

v1. Avoid memcpy in/out_syncobjs and fix some mistakes.
v2.
 - Handle vkQueueWaitIdle.
 - Add enum tu_semaphore_type.
 - Fix to handle VK_SEMAPHORE_WAIT_ANY_BIT_KHR correctly.
 - Fix a crash of dEQP-VK.synchronization.timeline_semaphore.device_host.misc.max_difference_value.

v3. Avoid indefinite waiting in vkQueueWaitIdle by calling
tu_device_submit_deferred_locked itself.

Signed-off-by: Hyunjun Ko <zzoon@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10126>

turnip: prep work for timeline semaphore support

Small refactor to classify semphore types, currently only binary
syncobj is being used though.

v1. Fix a crash of dEQP-VK.api.null_handle.destroy_semaphore

Signed-off-by: Hyunjun Ko <zzoon@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10126>

zink: pop descriptor refs when invalidating sets

this avoids memory ballooning, mainly for samplers, when a descriptorset
component has refs for a very, very large number of sets

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10699>

zink: remove screen param from zink_descriptors_update()

this was redundant

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10699>

zink: make zink_context_update_descriptor_states() static

this is no longer used externally

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10699>

zink: call invalidate on invalid descriptor sets during recycle

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10699>