ci/a3xx: Run all of GLES3 dEQP.

We're not spending half our time booting any more, so run the other half.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9314>

ci/a5xx: Run all of gles2 in one job.

Now that we're not spending so much time on boot overhead, no need to
parallelize.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9314>

ci/freedreno: Switch the fastboot boards to using nfsroot.

This saves time in packing the rootfs, allows for larger rootfses, and
avoids the need for webdav.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9314>

ci/freedreno: Also retest when only CI configuration changes.

Fixes: dab845d457eb ("ci: Move specific driver testing to separate files in separate dirs.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9314>

tgsi_exec: Jump over entirely non-taken THEN or ELSE branches.

TGSI has these nice labels for us for where to jump in this case, let's
use them.  Improves piglit arb_shader_image_load_store-shader-mem-barrier
runtime massively, though not enough to make the test really reasonable to
run.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9347>

tgsi_exec: Roll the loops for condmask handling.

No need to hand-unroll this, the compiler will do it.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9347>

i965: support GL_EXT_color_buffer_half_float

FP16 rendering is supported on all gen4 hardware.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9379>

ci: Use lock file to build deqp-runner

Build deqp-runner with the `--locked` option to use dependencies
versions specified in `Cargo.lock`.

v2: Bump image tags.

Signed-off-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Suggested-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9368>

ac/llvm: open code fpow on LLVM 12 using fmul.legacy

A quick look at the asm shows that this enables source modifiers
(neg, abs) for v_mul_legacy_f32.

Totals from affected shaders:
SGPRS: 110104 -> 110400 (0.27 %)
VGPRS: 57632 -> 57636 (0.01 %)
Spilled SGPRs: 66 -> 63 (-4.55 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 3290412 -> 3283068 (-0.22 %) bytes
Max Waves: 32141 -> 32141 (0.00 %)

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9395>

ac/llvm: add type parameter into ac_build_buffer_load to fix 16-bit TES inputs

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9395>

ac/llvm: fix visit_load_ubo_buffer to use SMEM for 16 bits instead of VMEM

This has 3 advantages:
- It's SMEM.
- Multiple single component loads are merged into 1 multi-dword load by LLVM.
- The result is always packed for packed instructions.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9395>

ac/llvm: implement 16-bit and 64-bit fpow correctly

LLVM converts to 32 bits and back for llvm.pow, so we can't use it.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9395>

ac/llvm: add support for 16-bit source operands for samplers

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9395>

nir/search: Constify instruction parameter to search helpers

The search helps must *never* modify the instruction passed in, so let
the compiler enforce this.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9378>

anv: fix missing general state pool in validation list

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 83fee30e85965c ("anv: allow multiple command buffers in anv_queue_submit")
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9388>

ci/lava: Move the driver expectation files to the per-driver CI dir.

This will cause less retesting of other drivers when changing the dEQP
results for a driver.

Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9353>

ci/lava: Move the per-driver gitlab-ci.yml to each driver.

Follow-up to !9139, will cause less testing of other drivers when changing
the CI configuration for a single driver.

Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9353>

ci: Move deqp-default-skips.txt back to .gitlab-ci/

Since we don't manually enumerate the drivers using it, we have to retest
all drivers when changing it (which basically never happens, anyway).

Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9353>

radv: fix RGP barrier layout transition for TC-compatible CMASK images

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/9390>

zink: Enable GL_EXT_depth_bounds_test

Available since Vulkan 1.0, and in fact already wired up, just not
advertised. It looks like we could make this dynamic state but this
works for now.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9371>

radv: don't shrink image stores for The Surge 2

The game seems to declare the wrong format.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: e4d75c22 ("nir/opt_shrink_vectors: shrink image stores using the format")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4347
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9229>

nir/opt_shrink_vectors: add option to skip shrinking image stores

Some games declare the wrong format, so we might want to disable this
optimization in that case.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: e4d75c22 ("nir/opt_shrink_vectors: shrink image stores using the format")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9229>

turnip: fix leak of tu_shader object during compute pipeline creation

tu_shader should be freed after pipeline is successfully created.

Fixes tests:
 dEQP-VK.api.object_management.alloc_callback_fail.compute_pipeline
 dEQP-VK.api.object_management.alloc_callback_fail_multiple.compute_pipeline

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9364>

radv: bump the initial SQTT buffer size to 32MB per SE

Most of the games need 32MB or more, but rarely less.

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/9367>

radv: trigger a new SQTT capture automatically after resizing the buffer

It's way better.

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/9367>

radv: double the SQTT buffer size when it is resized

Computing the expected buffer size isn't reliable on GFX10+ because
DROPPED_CNTR returns weird results.

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/9367>

ac/sqtt: fix determining if the trace is complete on GFX10+

DROPPED_CNTR isn't reliable and might still report non-zero if the
SQTT buffer isn't full. Checking if the number of written bytes by
the hw is equal to the SQTT buffer size seems reliable.

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/9367>

radv: do not trace inactive shader engines with SQTT

This fixes a GPU hang on my Sienna because the number of SE is
less than the maximum, and SE #1 is disabled.

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/9370>

zink: ci updates

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9291>

zink: use staging resource for write transfer_map in order to not stall

we can just give the user a staging resource and then flush the data back
later

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9291>

radeonsi: don't crash on NULL images in si_check_needs_implicit_sync

This fixes CTS test: KHR-GL46.arrays_of_arrays_gl.AtomicUsage

Fixes: bddc0e023c "radeonsi: fix read from compute / write from draw sync"

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9361>

ac/llvm: fix ac_build_atomic_rmw with LLVM 13

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4383

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9361>

nir/vec_to_movs: Don't generate MOVs for undef channels.

This appeared in softpipe's image operations, since NIR always uses
4-component values for the coords, while the GLSL IR only has 2 components
for a 2D image (for example).
arb_shader_image_load_store-shader-mem-barrier (which times out in CI and
spends its time inside of tgsi_exec) was spending 4/51 of its instructions
on moving these undefs around.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9345>

nir: Add a nir_src_is_undef() helper, like nir_src_is_const().

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9345>

zink: export PIPE_CAP_TGSI_VS_LAYER_VIEWPORT

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9283>

zink: enable PIPE_CAP_CLEAR_SCISSORED

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9283>

lavapipe: sort bindings before creating descriptor set

This ensures the dynamic offsets are correct

Fixes: b38879f8c5f5 ("vallium: initial import of the vulkan frontend")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9359>

lavapipe: reorder descriptor set stages to get correct binding

The fragment stage was in the wrong place here.

Fixes: b38879f8c5f5 ("vallium: initial import of the vulkan frontend")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9359>

gallium/dri: Remove dri2_format_mapping::cpp

I was suspicious that some entries in dri2_format_table (in
dri_helpers.c) had this field set incorrectly.  It seemed like
DRM_FORMAT_ABGR16161616F and DRM_FORMAT_XBGR16161616F should have been 8
instead of 4.  Upon digging I found that nothing uses the field.  Fix
code by removing it.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9354>

clover: Add missing include for llvm-12 build fix

Fixes: d1eab2b1eb1 ("clover: Fix build with llvm-12.")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9372>

zink: apply only the pending zs clear bits during deferred clears

both bits will have been flagged at this point in order to indicate
that the aspects will be cleared "at some point" during the loop, but
when actually iterating through the pending clears, only the bits set
in the clear call should be applied

Fixes: 5c629e9ff24 ("zink: defer pipe_context::clear calls when not currently in a renderpass")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9366>

st/nine: Simplify checks for driconf options

Remove the useless driCheckOption calls. They always
succeed.

As a result the intended behaviour for thread_submit
was not working (different default depending on the gpu
used). Add a comment to fix that in the future.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

driconf: Rename csmt_int back to csmt_force

Fixes regression introduced by
<https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6916>

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Fix leak at device destruction

At the release of the last object holding a reference
on the device, the device dtor was executed and
the objector dtor was ignored.

The proper way is to execute the object dtor, then
the device dtor.

The previous code was likely for a workaround against
something that was fixed since.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Protect *PrivateData also for Volumes

*PrivateData functions were not protected by
a mutex for Volumes whereas they definitely
should.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Refactor ht_guid_delete

Have ht_guid_delete take a hash_entry.

As a result, we can use _mesa_hash_table_remove instead of
_mesa_hash_table_remove_key.
The previous code using the latter was incorrect as the key
of the entry was read after it was freed.

Fixes: https://github.com/iXit/wine-nine-standalone/issues/40

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Add new debug and error checks

Add new debug messages and error checks

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Enable DF24 support

We can enable it, now that FETCH4 is
implemented.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Implement experimental FETCH4

FETCH4 is a feature that needs to be implemented
to advertise D3DFMT_DF24.
It's basically a variant of Gather4.

This first implementation will need to be completed
to implement the feature fully, but the feature
doesn't seem to be much used (other equivalent
features are preferred by games).

Note until DF24 is advertised, apps are not supposed to use
FETCH4.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Track formats compatible with FETCH4

FETCH4 is a d3d9 extension not much used, as newer
ones were prefered. However it's support is required
to advertise the DF24 format.

Prepares support by tracking compatible formats.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Unmap buffers after full unlock

Do not unmap anything until all buffer unlocks
were received.

A buffer can be filled in several threads, and thus
in the case of double locks, it's not possible to know
which unlock is received first.
Thus only unmap the buffers when the last unlock is
received.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Clamp GetAvailableTextureMem

Previously we used to clamp "available_texture_limit",
which was incorrect. "available_texture_mem" should
have been clamped instead.
The resulting code was noop.

The idea behind that code was that 32 bits executable
would see maximum 4GB video memory.

However it seems according to users that 32 bits apps
should be able to allocate more than 4GB, thus the
clamping is inappropriate.

Instead clamp the return of GetAvailableTextureMem, to
correctly report a high value when there is more than
4GB available.

I do not know what should exactly be the clamp value,
for now have a 64MB margin below UINT_MAX.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Do not allow depth buffer render targets

Without the proposed check, some apps will decide to use depth buffers
as render targets.

Bug found investigating:
https://github.com/iXit/wine-nine-standalone/issues/82

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

st/nine: Reduce system memory allocated by D3DUSAGE_AUTOGENMIPMAP

For D3DUSAGE_AUTOGENMIPMAP basically, everything behaves
for the application as if the texture had one level.
However the pipe_resource has more levels, and those
get generated automatically.

Previously we did allocate all the Surfaces as if
the texture had all the levels, except of just one.
The app could still just access the first level.

This patch completly removes the useless unaccessible
Surfaces.
In addition removes redundant handling of D3DUSAGE_AUTOGENMIPMAP.

Signed-off-by: Axel Davy <davyaxel0@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9177>

r600/sfn: Update status

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600: Enable GLSL 450 for nir shaders.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600: Don't optimize using source modifiers on literals

The code improvement is limited and it interferes with using literals
directly in LDS index ops, since here source modifiers are not
supported, but the current assembler code might inject the modifiers.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: Fix loading TES gl_PatchVerticesIn

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: handle querying the number of layers in cube arrays

This has to be loaded from a constant buffer instead of the actual
texture.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

nir: disaallow reordering for r600 shared load and remove component field

The original shared load op can't be reordered, so it might be better to
also not allow this for the lowered variant.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Acked-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: encode component in address for local IO

The backend code was actually assuming this, but the lowering still set
the components and write masks like it would be honoured.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: remove some old debug output

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: remove unused emit_alu_op2_split_src_mods

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: remove code for nir_op_fsign since it is lowered

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600: unify nir shader options evaluation

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: Allow any channel for the helper invocation evaluation

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: lower isign and iabs in nir

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: set info about using helper_invocation to skip sb

sb can't handle helper invocations, so skip sb when it is used.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

r600/sfn: Lower FS inputs to temps late and, and lower interpolate at

This fixes FS shaders where a var is loaded with two different
interpolators.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9330>

util: Always use timespec_get on Windows.

include/c11/threads_win32.h provides a fallback implementation of
timespec_get when necessary.

Fixes https://gitlab.freedesktop.org/mesa/mesa/-/issues/4109

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9280>

aco: add missing usable_read2 check

A Hitman 2 shader does: read64(local_invocation_index() * 4 - 4). This was
likely emitting a ds_read2_b32 on GFX6. For local_invocation_index()=0,
because the first dword was out-of-bounds, the second was likely also
considered out-of-bounds (even though it's not, at offset 0).

Likely fixes https://gitlab.freedesktop.org/mesa/mesa/-/issues/3882

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 57e6886f981 ("aco: refactor load_lds to use new helpers")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9332>

Revert "radv,aco: allow unaligned LDS access on GFX9+"

This reverts commit 1a0b0e8460c1881f94975b3dfbe9c312d9c3fcf7.

The bounds checking behaviour of ds_read_b64, ds_read_b96 and ds_read_b128
make this feature very difficult to use safely.

This fixes a blocking artifact in Hitman 2. Previously, it contained:
ds_read_b64(local_invocation_index() * 4 - 4)
For local_invocation_index()=0, the second dword would be considered
out-of-bounds, even though it's at offset 0.

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9332>

broadcom/compiler: ldvary pipelining tracking and documentation clean-ups

Now that we can pipeline all varyings we should not be referring
specifically to smooth varyings anywhere.

Also, rename the instruction field 'ldvary_pipelining' to
'is_ldvary_sequence', which is more appropriate, since we always
set this for any instruction involved with varying setups,
independently of whether they end up being pipelined or not.

This also does some other minor edits which intend to slightly
simplify the code and make it a bit more compact.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9363>

glsl/float64: Bump #version to 400

An earlier commit tried to make this shader compatible with GLSL 3.30,
but it requires, GL_ARB_gpu_shader_int64, which requires GLSL 4.00 and
GL 4.0 according to the extension spec.  So we were failing to enable
the required extension, breaking compilation of this shader.

The original intention of that patch was to get this working on zink,
which at the time only supported GL 3.3.  But now it supports later
OpenGL versions, so we don't need to do this any longer.  Rather than
revert the patch and raise the version all the way back to 430, just
bump it to the require 400 at Ian Romanick's suggestion.

Fixes: 4d47b22bf00 ("glsl/float64: make this compatible with glsl 330")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3991
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9351>

clover: Fix build with llvm-12.

Fix build error after LLVM commit c495dfe0268b ("[clang][cli] NFC:
Decrease the scope of ParseLangArgs parameters").

../src/gallium/frontends/clover/llvm/invocation.cpp: In function ‘std::unique_ptr<clang::CompilerInstance> {anonymous}::create_compiler_instance(const clover::device&, const string&, const std::vector<std::__cxx11::basic_string<char> >&, std::string&)’:
../src/gallium/frontends/clover/llvm/invocation.cpp:252:55: error: cannot convert ‘clang::PreprocessorOptions’ to ‘std::vector<std::__cxx11::basic_string<char> >&’
  252 |                                 c->getPreprocessorOpts(),
      |                                 ~~~~~~~~~~~~~~~~~~~~~~^~
      |                                                       |
      |                                                       clang::PreprocessorOptions

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4114
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8543>

broadcom/compiler: allow pipelining of flat and noperspective varyings

These end up having a NOP between the ldvary and the next instruction
in the sequence (a MOV for flat and an FADD for noperspetive):

nop                  ; nop               ; ldvary.r0
nop                  ; nop
fadd  rf6, r0, r5    ; nop               ; ldvary.r1
nop                  ; nop
fadd  rf5, r1, r5    ; nop               ; ldvary.r2
nop                  ; nop
fadd  rf4, r2, r5    ; nop               ; ldvary.r3

To pipeline these, we can reuse the same infrastructure we have in
place for smooth varyings but we need to avoid breaking the sequence
due to the NOP instruction. We do that by testing if dropping the
sequence when we failed to pick up the next instruction also fails
to choose an instruction.

This is not perfect, because we may be able to choose an instruction
outside the sequence such as an ldunif, and use that to break a
sequence that we could otherwise continue after scheduling the NOP
instruction, but it is still better than nothing.

total instructions in shared programs: 13820690 -> 13819774 (<.01%)
instructions in affected programs: 64026 -> 63110 (-1.43%)
helped: 479
HURT: 62
Instructions are helped.

total max-temps in shared programs: 2326435 -> 2326423 (<.01%)
max-temps in affected programs: 102 -> 90 (-11.76%)
helped: 7
HURT: 0
Max-temps are helped.

total sfu-stalls in shared programs: 30683 -> 30710 (0.09%)
sfu-stalls in affected programs: 13 -> 40 (207.69%)
helped: 2
HURT: 24
Sfu-stalls are HURT.

total inst-and-stalls in shared programs: 13851373 -> 13850484 (<.01%)
inst-and-stalls in affected programs: 62818 -> 61929 (-1.42%)
helped: 466
HURT: 65
Inst-and-stalls are helped.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9304>

broadcom/compiler: pipeline smooth ldvary sequences

Typically, we would schedule smooth varyings like this:

nop                  ; nop               ; ldvary.r4
nop                  ; fmul  r0, r4, rf0
fadd  rf13, r0, r5   ; nop               ; ldvary.r1
nop                  ; fmul  r2, r1, rf0
fadd  rf12, r2, r5   ; nop               ; ldvary.r3
nop                  ; fmul  r4, r3, rf0
fadd  rf11, r4, r5   ; nop               ; ldvary.r0

where we pair up an ldvary with the fadd of the previous sequence
instead of the previous fmul. This is because ldvary has an implicit
write to r5 which is read by the fadd of the previous sequence, so
our dependency tracking doesn't allow us to move the ldvary before the
fadd, however, the r5 write of the ldvary instruction happens in the
instruction after it is emitted so we can actually move it to the fmul
and the r5 write would still happen in the same instruction as the fadd,
which is fine.

This patch allows us to pipeline these sequences optimally. For that,
after merging an ldvary into a previous instruction in the middle of
a pipelineable ldvary sequence, we check if we can manually move it
to the last scheduled instruction instead (the one before the
instruction we are currently scheduling).

If we are successful at moving the ldvary to the previous instruction,
then we flag the ldvary as scheduled immediately, which may promote
its children (the follow-up fmul instruction for that ldvary) to DAG
heads and continue the merge loop so that fmul can be picked and
merged into the final fadd of the previous sequence (where we had
originally merged the ldvary). This leads to a result that looks like
this:

nop                  ; nop               ; ldvary.r4
nop                  ; fmul  r0, r4, rf0 ; ldvary.r1
fadd  rf13, r0, r5   ; fmul  r2, r1, rf0 ; ldvary.r3
fadd  rf12, r2, r5   ; fmul  r4, r3, rf0 ; ldvary.r0

Shader-db results:

total instructions in shared programs: 14071591 -> 13820690 (-1.78%)
instructions in affected programs: 7809692 -> 7558791 (-3.21%)
helped: 41209
HURT: 4528
Instructions are helped.

total max-temps in shared programs: 2335784 -> 2326435 (-0.40%)
max-temps in affected programs: 84302 -> 74953 (-11.09%)
helped: 4561
HURT: 293
Max-temps are helped.

total sfu-stalls in shared programs: 31537 -> 30683 (-2.71%)
sfu-stalls in affected programs: 3551 -> 2697 (-24.05%)
helped: 1713
HURT: 750
Sfu-stalls are helped.

total inst-and-stalls in shared programs: 14103128 -> 13851373 (-1.79%)
inst-and-stalls in affected programs: 7820726 -> 7568971 (-3.22%)
helped: 41411
HURT: 4535
Inst-and-stalls are helped.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9304>

broadcom/compiler: track pipelineable ldvary sequences

If we have two (or more) smooth varyings like this:

nop t3; ldvary.rf0
fmul t5, t3, t0
fadd t6, t5, r5
nop t7; ldvary.rf0
fmul t9, t7, t0
fadd t10, t9, r5
nop t11; ldvary.rf0
fmul t13, t11, t0
fadd t14, t13, r5

We may be able to pipeline them like this:

nop                  ; nop               ; ldvary.r4
nop                  ; fmul  r0, r4, rf0 ; ldvary.r1
fadd  rf13, r0, r5   ; fmul  r2, r1, rf0 ; ldvary.r3
fadd  rf12, r2, r5   ; fmul  r4, r3, rf0 ; ldvary.r0

But in order to do this, we will need to manually tweak the
QPU scheduling.

This patch tracks information about ldvary sequences that are
good candidates for pipelining, and a follow-up patch will
use this information to pipeline them when we emit the QPU
code.

v2 (apinheiro):
  - Rename the v3d_compile fields to avoid confusion with the qinst fields.
  - Assert that a sequence's start instruction is not the same as the end.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9304>

broadcom/compiler: fix indentation style

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9304>

broadcom/compiler: fix DAG pre-remove for merged instructions

When selecting an instruction to merge, we want to pre-remove that
instruction from the DAG, not the one we are merging it in, which
we had already pre-removed right before.

The reason this was not causing problems before is that the
consequence of this bug is we will choose the same instruction
again in the merge loop and trying to merge that instruction twice
will fail and we would break out of the merge loop and move on.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9304>

anv: toggle on sample shading if it is set in the shader

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4300
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9040>

intel: Use devinfo genx10 field

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9329>

intel/dev: Add devinfo genx10 field

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9329>

intel: Use GEN_VERSIONx10 in more places

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9329>

iris: fix in fences backend for ext_external_objects edge case

EXT_external_objects require we call glSignalSemaphoreEXT followed by a
glFlush. If the rendering workload is small when Signal and Flush
take place the relevant batch buffers with the actual rendering might
have been submitted already. In that case the following condition is met:
(iris_batch_bytes_used(batch) == 0). This causes:
glFlush() --> iris_fence_flush() -> iris_batch_flush() ->
_iris_batch_flush() to no-op, and so the fence doesn't get submitted to the
kernel. Then when anv tries to submit an execuf2 that must wait on the
shared VkSempahore / drm_syncobj fence, there isn't one and the kernel
rejects the batchbuffer causing an -EINVAL return of the execbuf2 ioctl
and a VK_DEVICE_LOST error. Empty batch buffers do have typically one
fence attached, but the ones carrying the extra fence from a
glSignalSempahore() call do have at least 2.

See also: the discussion in MR!4337.

v2: Changed the batch struct to have a contains_fence_signal variable
that is set to true when i915_EXEC_FENCE_SIGNAL fence is added to the
batch and off when batch is reset (Tapani Pälli)

Authored-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reported-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8861>

util/disk_cache: move cache tests to the util directory

The cache has been detangled from glsl and used outside it (with Vulkan drivers)
for years now.

This also cleans up the dependancies in the build file. The test doesn't
depend on the glsl lib but rather the util lib.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9327>

lavapipe: VK_EXT_extended_dynamic_state support

zink can use this in some paths, and it's not a crazy thing to support
for lavapipe.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9328>

radeonsi: don't index si_context::shaders with enum gl_shader_stage

Fixes: a8373b3d387 "radeonsi: store si_context::xxx_shader members in union"

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9313>

lavapipe: Define supported extensions in C

One less python generator, like anv/radv

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9075>

lavapipe: expose a 1.0 vulkan API for now.

lavapipe doesn't have all 1.1 features yet, just return
the baseline 1.0 API always.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9075>

zink/instance: work with vulkan 1.0 and later loader.

If zink is meant to work against Vulkan 1.0 API then it
should expose the 1.0 API as create time as well as always
ask for all the vulkan 1.0 extensions.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9075>

mesa/st: even better no-oping for clears

<imirkin> zmike: re 6a8c51dc5a4 -- couldn't minx be > maxx after the clamping? in nouveau i did >= for the bail condition

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9350>

zink: ci updates

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9276>

zink: handle GLSL_SAMPLER_DIM_EXTERNAL in ntv

this fixes most of the tests that use it

Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9276>

tnl: Call _mesa_matrix_analyse to make sure the inverse MVP is updated

A recent commit stopped updating the inverse MVP matrix, because usually
only GLSL built-ins need it.  However, TNL also needs it.  So make sure
it's correct when needed.

Fixes: 10371c520c1 ("mesa: don't compute the ModelView * Projection matrix if not used")
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9346>

android: driconf: Generate a static table when no xmlconfig

Fixes the following building error:

FAILED: out/target/product/x86_64/obj_x86/STATIC_LIBRARIES/libmesa_util_intermediates/xmlconfig.o
...
external/mesa/src/util/xmlconfig.c:1030:12: fatal error: 'driconf_static.h' file not found
           ^~~~~~~~~~~~~~~~~~
1 error generated.

Fixes: a6b0ceb ("driconf: Generate a static table when no xmlconfig")
Acked-by: Rob Clark <robclark@freedesktop.org>
Tested-by: Marijn Suijten <marijn.suijten@somainline.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9294>

driconf: avoid Non-ASCII character error in driconf_static.py

MR to move to python3 in Android build gen rules is still pending

The change is to avoid following building error:

FAILED: out/target/product/x86_64/gen/STATIC_LIBRARIES/libmesa_util_intermediates/driconf_static.h
/bin/bash -c "/usr/bin/python external/mesa/src/util/driconf_static.py external/mesa/src/util/00-mesa-defaults.conf > out/target/product/x86_64/gen/STATIC_LIBRARIES/libmesa_util_intermediates/driconf_static.h"
  File "external/mesa/src/util/driconf_static.py", line 2
SyntaxError: Non-ASCII character '\xc2' in file external/mesa/src/util/driconf_static.py on line 2, but no encoding declared; see http://python.org/dev/peps/pep-0263/ for details

Fixes: a6b0ceb ("driconf: Generate a static table when no xmlconfig")
Acked-by: Rob Clark <robclark@freedesktop.org>
Tested-by: Marijn Suijten <marijn.suijten@somainline.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9294>

d3d12: Handle null constant buffers

RuneScape ends up hitting this path, and it's easy enough to get
some well-defined behavior instead of a crash.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-By: Bill Kristiansen <billkris@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9342>

mesa/st: no-op scissored clear calls with size zero

Wx0 and 0xH should result in no-ops in the driver, so they can just
become no-ops before they reach the driver to save some validation later

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9344>

mesa/st: clamp scissored clear regions to fb size

these should never be larger than the fb and drivers shouldn't have to
care about it

Fixes: 1c8bcad81a7 ("gallium: add pipe cap for scissored clears and pass scissor state to clear() hook")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9344>