mesa: Remove unused Mesa IR PROGRAM_* register files.

Now that glsl-to-tgsi is gone, a bunch of this is dead.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

glsl: Remove UBO reference lowering.

All UBO-supporting drivers now go through the NIR path, which does a
better job of it.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

tgsi_scan: Remove unused input_array_last/output_array_last.

Last use removed in "radeonsi: remove TGSI" this year.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

mesa/st: Remove TGSI-only shader lowering code.

These are no longer called.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

mesa/st: Drop the attrib remapping in st_prepare_vertex_program().

This was only used in the GLSL-to-TGSI path, which is no longer.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

mesa/st: Remove TGSI shader program variant handling.

We no longer have anything generating gl_programs with TGSI in them since
retiring GLSL-to-TGSI.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

mesa/st: Remove st_glsl_to_tgsi.

It is no longer called, and can be retired.

Fixes: #1924, #822, #6073

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

mesa/st: Always generate NIR from GLSL, and use nir_to_tgsi for TGSI drivers.

The NIR path through the frontend is effectively the only one maintained
for a quite a while now.  We can see that effect with !15540, where the
TGSI generation path was regressed to assertion fail on real-world
shaders, and nobody noticed until I came along trying to test the
NIR-to-TGSI transition.

We already have a nir_to_tgsi() call for translating NIR representation
for ARB programs into TGSI before handing them off to the driver.  This
change makes that path get taken for GLSL programs as well.

This is the minimum change to get all the drivers on NIR from GLSL, to
give a simple commit to bisect too.  The dead code removal comes next.

Now every driver benefits from shared NIR optimizations for GLSL, and we
can start retiring GLSL optimizations.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

ci/iris: Add a bunch of APL and KBL flakes recently.

I got hit by one of them trying to merge !8044.  Just update the list.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8044>

kopper: pass the current context to dri_flush

passing the drawable's context leads to desync and crashing if the app
is using multiple threads and multiple contexts

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16351>

kopper: invalidate drawables when resizing textures in place

this guarantees that swapchains shared between contexts will stay
in sync

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16351>

aco: ensure that definitions fixed to operands have matching regclasses

If the operand is not killed, the definition needs to be large enough so
that the new location for the operand does not intersect with the old
location.

Fixes with zink:
KHR-GL45.shader_image_load_store.basic-allTargets-atomicCS
KHR-GL45.shader_image_load_store.basic-allTargets-atomicGS
KHR-GL45.shader_image_load_store.basic-allTargets-atomicVS

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6276
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16106>

ci/iris: Demote APL deqp to manual-only for now.

it's been flaking with "2022-05-05 16:29:49.055151: [0m[31mERROR - Failure
getting run results: parsing results: Reading from dEQP: timed out waiting
for fd to be ready (See \"//results/c32.r1.log\")" and a pile of missings
since the brief "whoops, HW CI failed to listen to the test exit code"
regression.

The only ways I know of to hit this case would be:

1) The deqp binary abruptly wedges on its own.  This happens with NFS
failures sometimes, but the rest of the run went fine and we never got the
kernel complaining about NFS, so that seems unlikely.

2) The stderr pipe filled up before stdout was completed, and deqp got
wedged trying to output stderr (happens sometimes when you do like
NIR_DEBUG=print in your run).

Both of these seem unlikely, given that we've got a big .qpa file that
made it all the way to writing out test case durations at the end of the
run before abruptly terminating.  Why didn't we have at least some of the
test results parsed?

The next deqp-runner release we integrate will solve #2, and cleans up
these error paths a bunch, so I'm hoping we get more information soon.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16350>

panfrost: Increase the maximum tiler heap size

Required to avoid tiler heap out-of-memory condition on Valhall on tests
including:

dEQP-GLES31.functional.draw_indirect.compute_interop.large.drawelements_combined_grid_1200x1200_drawcount_8

This test passes on Bifrost without the fix because varyings are only allocated
from the tiler heap on Valhall.

Minimal perf or memory usage impacted is expected, as even old versions of
panfrost.ko support growable memory.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16330>

docs/features: flag VK_KHR_timeline_semaphore as implemented

We got this implemented for v3dv when Jason ported the submit code
to the common submit framework.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16344>

v3dv: trivially implement VK_KHR_separate_depth_stencil_layouts

We don't currently benefit from seeing barriers and layout
transitions that affect just the depth or stencil aspects,
so we don't expose this feature.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16344>

v3dv: implement VK_EXT_separate_stencil_usage

We don't care about the depth/stencil usage other than to determine
if an image format is supported for that usage.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16344>

v3dv: check input attachment usage as sampled usage

Since we implement input attachments as textures we should check
support for input attachment usage the same way we check support
for sampled images.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16344>

pvr: Update pvrsrvkm to fw 1.17 .

Signed-off-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16238>

v3dv: use wait stage for wait semaphores to decide about binning syncs

If the wait stage for a semaphore doesn't involve geometry stages we
don't need a binning sync.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16322>

v3dv: skip binning sync if binning shaders don't access external resources

Until now we have been enabling binning sync if we found a barrier
involving geometry stages (a bcl barrier), however, if the actual
binning shaders involved with the job don't access any external
buffers or images there is no reason to sync at the binning stage.

In this patch we don't immediately consume the bcl barrier flag from
the command buffer state when we create a new job. Instead, we check
this state when we are about to emit a draw call by checking if the
shaders involved with binning may access external resources, such as
vertex buffers, UBOs, or textures. If none of the draw calls in the
job use binning shaders that access external resources then we never
enable binning sync for the job.

It is possible that a binning shader uses resources that are not
synchronized through a barrier though, so we keep track of the
access masks used with barriers for both buffers and images separately
to better identify if the binning shader is affected by the barrier.

If a serialized job never consumes the bcl barrier flag because none
of its draw calls ever required a bcl sync, then the flag will be
cleared when the job is finished.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16322>

v3dv: flag BCL barriers for all-graphics and all-commands stages

Since these also include geometry stages in the pipeline.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16322>

v3dv: drop default sampler states if not used

If a shader doesn't use any samplers (including default sampler states),
make sure we drop them so other parts of the driver can recognize that
the program doesn't actually use any samplers at all.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16322>

radv: disable instance packing to fix pipeline query stats

RDNA2 is affected by a hardware bug when instance packing is enabled
for adjacent primitive topologies and instance_count > 1, pipeline
stats generated by GE are incorrect. It needs to be applied for
indexed and non-indexed draws.

This is based on PAL waDisableInstancePacking.

This fixes KHR-GL46.pipeline_statistics_query_tests_ARB.* with Zink.

Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6257
Cc: mesa-stable
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/15877>

ci: Fix tests expectations

For some days, the CI was bypassing LAVA and bare-metal jobs due to an
issue in the init-stage2.sh script. After the fix some tests
crashed/failed. This commit updates the expectations for them.

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16325>

ci: Update trace after CI fix

For some days, the CI was bypassing LAVA and bare-metal jobs due to an
issue in the init-stage2.sh script. After the fix the neverball trace on
panfrost-t860 is producing a different image, due to a bugfix in
Mesa itself driver. This commit updates the neverball trace on that
device.

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16325>

ci: Fix init-stage2 exit code

After a LAVA job submitter rework, the init-stage2.sh was changed to be
compatible with new LAVA job definitions, but the result from the script
represented by `HWCI_TEST_SCRIPT` variable is obfuscated by the `set -e`
command. So when the test script fails, `set` will override the exit
code and the jobs will pass when they should fail.

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16325>

docs: Add sh256sum for mesa 20.0.3

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

docs: add release notes for 22.0.3

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

docs: update calendar for 22.1.0-rc4

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

docs: update calendar for 22.1.0-rc3

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

docs: Extend calendar entries for 22.1 by 1 releases.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

docs: update calendar and link releases notes for 22.0.3

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

docs: Extend calendar entries for 22.0 by 1 releases.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16336>

zink: Fix memory leak on error path.

Fix defect reported by Coverity Scan.

Resource leak (RESOURCE_LEAK)
leaked_storage: Variable sampler_view going out of scope leaks the storage it points to.

Fixes: 625457caafb ("zink: handle swapchain acquire failures more directly")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16258>

vulkan/wsi: Add signal_fence/semaphore_for_image helpers

These operations are about to get a bit more complex so let's add a
couple nice helpers to keep things clean.

Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16333>

vulkan/wsi: Set the right stage flags for semaphore waits

This is currently technically broken for compute.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Cc: mesa-stable@lists.freedesktop.org
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16333>

vulkan/wsi: Reset the image fence right before vkQueueSubmit

Instead of resetting at the top of the loop, we create the fence in the
signaled state and then unconditionally reset it before vkQueueSubmit.
This gives us a bit more flexibility with how we handle the fence in the
case where the client provides zero semaphores.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16333>

vulkan/wsi: Hang on to file descriptors

Instead of closing the dma-buf file descriptors immediately after
handing them to the window system, hang on to them.  We want to be able
to use them for synchronization.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16333>

vulkan/wsi: Only use a single fd per wsi_image

The only thing this was helping was X11 where the protocol requires that
we pass in an array of images.  We can move all the dup() code to the
X11 back-end and leave the others a bit cleaner.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16333>

vulkan/wsi: Close file descriptors in wsi_destroy_image

Most of the time, this is a non-issue because the WSI back-end closes
them as part of handing them to the window-system and sets fds[*] to -1.
The one exception here was Wayland which was closing them but leaving
fds[*] pointing to bogus file descriptors.  Having wsi_destroy_image
close them makes clean-up easier and more reliable.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16333>

gallium: refactor a channel loop in draw_llvm.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: LLVM opaque pointers small changes

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_tgsi_aos.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_tgsi_soa.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_struct.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_sample_soa.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_sample.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_nir_soa.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_gather.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_format_s3tc.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_intr.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_flow.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_printf.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_coro.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_conv.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_format_aos.c

Also extract lp_build_const_func_pointer_from_type() in lp_bld_const.h
taking explicit function type.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_assert.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallivm: use LLVM opaque pointers in lp_bld_arit.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallium/llvmpipe: use LLVM opaque pointers in lp_bld_depth.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallium/llvmpipe: use LLVM opaque pointers in lp_bld_interp.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallium: use LLVM opaque pointers in draw_llvm.c

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

gallium: add opaque pointers shim for LLVM < 8.0

LLVM is transitioning to "opaque pointers", and as such deprecates
LLVMBuildGEP, LLVMBuildCall, LLVMBuildLoad, replacing them with
LLVMBuildGEP2, LLVMBuildCall2, LLVMBuildLoad2 respectivelly.
These new functions were added in LLVM 8.0; so for LLVM before 8.0 we
simply forward to the non-opaque-pointer variants.

Acked-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15893>

wsi/x11: Hook up KHR_incremental_present

We create one XFIXES region per swapchain image. If the QueuePresent
comes in with a list of rectangles, we push them into the region and
pass it to xcb_present_pixmap.

The extension is technically just a hint. We still fall back to the
unhinted "update the whole image" path if the update region has more
than an arbitrary number of rects, or if we're stuck using plain
PutImage instead of ShmPutImage.

Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16218>

zink: add a ci flake

weird that this is suddenly failing on two drivers lately...

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16329>

lavapipe: lower quad_broadcast intrinsics

not supported by gallivm

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16301>

svga: Add support for requesting NIR and translating to TGSI.

I'm working on switching mesa/st to no longer produce TGSI on its own, and
so we need a way to test SVGA against that future.

Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14160>

intel/dev: add RPL-S pci ids from drm-intel-next

from Tejas Upadhyay 'drm/i915: Add RPL-S PCI IDs' in drm-intel-next

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16320>

radv: only init acceleration structure if RT is enabled

This is to fix a LLVM crash with 13.0 because ATOMIC_FMAX is only
supported on 14.0+, so RADV_DEBUG=llvm was just completely broken.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16305>

radv/sqtt: fix configuring AUTO_FLUSH_MODE on GFX10.3

The polarity is inverted. Ported from RadeonSI and PAL.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16303>

ac,radeonsi: add has_sqtt_auto_flush_mode_bug

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16303>

egl/kopper: Hook up eglSwapInterval

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15800>

glx/kopper: Enable GLX_EXT_swap_control etc.

This requires newly tracking the max swap interval since kopper can't do
abs(interval) > 1 yet.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15800>

kopper: Grow a swap interval API

We take a slight liberty here by allowing 0 to mean either MAILBOX or
IMMEDIATE, since Wayland (at least) doesn't have a true IMMEDIATE mode
at least MAILBOX won't throttle to vblank.

This only correctly handles intervals of 0 or 1 at the moment.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15800>

wsi/x11: Avoid using xcb_wait_for_special_event in FIFO modes

If the window is destroyed from underneath us while we happen to be in
xcb_wait_for_special_event, there's no recovery. The special event will
never match because the XID is no longer valid, and Present doesn't have
an in-band DestroyNotify. We're going to work around this by using the
poll API instead. If we get an event we short-circuit back to the top of
the "wait for available image" loop, so we drain the whole special event
queue before any other logic. Which means if we run out of special
events (and the connection and swapchain are still valid) that we
_don't_ have enough images available, so to hurry along any events that
the X server hasn't flushed out yet we call GetGeometry on the
swapchain's window. As a side effect this verifies that the window is
still alive.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15800>

radv/radix: handle intentional allocation failures properly

This test can intentionally make the alloc callback to return NULL, so
we have to handle object creation failures properly. The driver should
also avoid memleaks because the test checks that.

Fixes crashes with
dEQP-VK.api.device_init.create_instance_device_intentional_alloc_fail.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16298>

radv: Use RADV_RT_STAGE_BITS more often

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16306>

radv: Treat rt stages like compute stages

Fixes dEQP-VK.binding_model.descriptorset_random.sets4.noarray.ubolimitlow.sbolimitlow.sampledimglow.outimgtexlow.noiub.nouab.rgen.noia.0
and probably some other ones.

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16306>

radv: Move RADV_RT_STAGE_BITS to radv_private.h

Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16306>

radv: re-emit dynamic line stipple state if the primitive topology changed

The dynamic primitive topology could change from LINE_LIST to
LINE_STRIP for example and the stipple state depends on this.

Found by inspection.

Cc: mesa-stable
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/16295>

pan/bi: Rework varying linking on Valhall

Valhall introduces hardware-allocated varyings. Instead of allocating varying
descriptors on the CPU with a slot based interface, the driver just tells the
hardware how many bytes to allocate per vertex and loads/stores with byte
offsets. This is much nicer!

However, this requires us to rework our linking code to account for separable
shaders. With separable shaders, we can't rely on driver_location matching
between stages, and unlike on Midgard, we can't resolve the differences with
curated command stream descriptors. However, we *can* rely on slots matching. So
we should "just" determine the byte offsets based on the slot, and then
separable shaders work.

For GLES, it really is that easy.

For desktop GL, it's not -- desktop GL brings unpredictable extra varyings like
COL1 and TEX2. Allocating space for all of these unconditionally would hamper
performance. To cope, we key fragment shaders to the set of non-GLES varyings
written by the linked vertex shader. Then we may define an efficient ABI, where
only apps only pay for what they use.

Fixes various tests in dEQP-GLES31.functional.separate_shader.random.* on
Valhall.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16310>

panvk: Don't use VARYING_SLOT_TEX0 internally

This is a legacy varying for desktop GL use. Don't use it in our meta shaders,
as it adds pointless complexity.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16310>

panfrost: Don't use VARYING_SLOT_TEX0 internally

This is a legacy varying for desktop GL use. Don't use it in our internal
shaders, as it adds pointless complexity.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16310>

panfrost/ci: Smoke test spilling

Spilling is tricky and doesn't get much testing in CI. Run
a subset of dEQP-GLES2.functional.shaders.* with spilling forced to get spilling
tested in CI.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

panfrost: Use packed TLS on Valhall

Packed TLS has cache-locality benefits on Valhall, compared to Bifrost's flat
TLS. Valhall does support flat TLS, but requires extra arithmetic in the shader
for correct results. At least until we get to generic pointers (and maybe even
then), we can use packed TLS. So just use packed TLS always for proper spilling.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

panfrost: Use emit_tls

Instead of rolling our own, so it can pick up a Valhall fix.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

panfrost: Correct XML for TLS

It was never updated for Valhall, from Midgard.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

pan/bi: Assert that blend shaders may not spill

The set of blend shaders is closed. They are completely internal. As such, we
know that the registers we reserve for them suffice, and we don't permit
register spilling. Refusing to support spilling in blend shaders simplifies a
number of parts of the compiler. Add a check that we don't try to spill anyway,
which will silently fail.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

pan/bi: Add option to test spilling

BIFROST_MESA_DEBUG=spill now restricts the register file to 1/4 its usual size,
useful for testing register spilling (e.g. running CTS) as well as debugging
spilling on small shaders.

Note blend shaders are exempt, as we don't allow blend shaders to spill.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

pan/bi: Align spilled registers on Valhall

Required to support packed addressing correctly. Fixes (with spilling forced):

dEQP-GLES2.functional.shaders.random.trigonometric.vertex.20

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

pan/va: Serialize memory stores

We could do better :(

Fixes spilling.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16314>

panfrost: Relax image check

Shader images on Valhall don't allow nonzero "Minimum level". However,
pan_texture lowers away nonzero minimum levels anyway, so there's nothing to
check. Fixes:

KHR-GLES31.core.shader_image_load_store.advanced-allMips-cs

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16315>

meson: Tell glslang to be quiet.

Without --quiet glslang unconditionally prints the input file name to stdout.
Check if --quiet is supported because some distros only have ancient glslang
versions.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16312>

aco/ra: fix vgpr_limit

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: b98a4d4dd7e ("aco: refactor GPR limit calculation")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16297>

aco: Remove D16 zero components from image stores.

No foz-db changes.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15179>

radv: Run copy_prop and dce after folding 16bit sampling/load/store.

Totals from 10 (0.01% of 134913) affected shaders:
CodeSize: 53168 -> 54832 (+3.13%); split: -0.17%, +3.30%
Instrs: 9117 -> 9200 (+0.91%); split: -1.74%, +2.65%
Latency: 41595 -> 41787 (+0.46%); split: -0.95%, +1.41%
InvThroughput: 16412 -> 16424 (+0.07%); split: -1.95%, +2.02%
VClause: 107 -> 112 (+4.67%); split: -0.93%, +5.61%
Copies: 199 -> 535 (+168.84%); split: -3.02%, +171.86%
PreVGPRs: 520 -> 502 (-3.46%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15179>

radv: Use nir_fold_16bit_image_load_store_conversions.

Totals from 10 (0.01% of 134913) affected shaders:
CodeSize: 53316 -> 53168 (-0.28%)
Instrs: 9219 -> 9117 (-1.11%)
Latency: 41744 -> 41595 (-0.36%)
InvThroughput: 16616 -> 16412 (-1.23%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15179>

aco: Implement image_load d16.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15179>

aco: Implement image_store d16.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15179>

nir/lower_mediump: Add a new pass to fold 16bit image load/store.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15179>