panfrost: Inline flip_compare_func into pan_encoder.h

This will become the home for little GenXML-aware helpers, suitable to
be #include'd from pan_cmdstream.c (or panvk equivalent).

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11785>

lavapipe: expose strict-lines feature

The strictLines-feature requires lines to be rasterized as rectangles
by default instead of using the parallelograms you get from extending
bresenham lines along their minor axis.

Now that we can specify the line mode fully we can actually express
this, so let's do so.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

lavapipe: re-expose line-rasterization extension

While we're at it, let's also enable rectangular line support, now that
we can properly control it.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

llvmpipe: respect rectangular_lines

With the new rectangular_lines state, we can now support rasterizing
wide lines correctly according to the vulkan spec, where this can be
specified independently of the rest of the state.

Because rectangular lines are orthogonal to multi-sampling, we now need
to also adjust with the pixel-offset in the rectangle code-path as well.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

draw: respect line_rectangular state

Now that we have a separate flag to signal rectangular lines, let's use
that instead of the smooth-flag.

This should have the benefit of also drawing rectangular lines when
multisampling stippled, non-smooth lines. But it seems there were no
test for that, so no results to update.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

gallium: explicitly specify line rasterization mode

Currently, drivers infer the line rasterization mode from the
multisampling and line_smooth rasterization state. This is always
correct for OpenGL, but is subtly incorrect for DirectX 9, 10 and
Vulkan (when VK_EXT_line_rasterization is supported).

So let's allow front-ends to choose freely between rectangle and
paralellogram rendering.

The reason why there's no added cap for this, is that the implicit
selection that drivers currently do will work just as well (or more
correclty, just as subtly wrong) as before. And there's nothing
reasonable the front-ends can do to get the correct behavior, so
there's really no fall-back code to write either.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

lavapipe: fix disable_multisample condition

There's two of the line-modes that warrants disabling multisampling,
and that's bresenham and smooth lines.

The reason we need this for the smooth lines case, is that multisampling
overrides the smooth-flag.

Fixes: 9fbf6b2abff ("lavapipe: implement VK_EXT_line_rasterization")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

lavapipe: do not disable multisampling for smooth lines

Smooth lines are also rasterized with rectangles instead of using
bresenham-style lines, so let's make sure we exclude those as well
from this test.

Fixes: 9fbf6b2abff ("lavapipe: implement VK_EXT_line_rasterization")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11782>

aco/meson: remove inc_gallium from include_directories

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11258>

aco: Format.

Manually adjusted some comments for more intuitive line breaks.

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11258>

aco: add .clang-format file

Based on src/amd/.clang-format with following changes:
Language: Cpp
Standard: c++14
PointerAlignment: Left
IndentAccessModifiers: False
SpaceBeforeCtorInitializerColon: True
AllowShortCaseLabelsOnASingleLine: True
AllowShortFunctionsOnASingleLine: InlineOnly
AllowShortBlocksOnASingleLine: Empty
AllowShortLambdasOnASingleLine: All
BraceWrapping:
  SplitEmptyFunction: false
  SplitEmptyRecord: false
  BeforeLambdaBody: true
IncludeBlocks: Regroup (with specified IncludeCategories)

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11258>

zink: ci updates

this one no longer flakes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11829>

zink: zero out sampler/image descriptor surface info for null descriptor updates

this is the surface struct used by the caching manager to generate a ref update template,
which means it must be kept in sync with the current state of descriptors

when the struct is zeroed, it implies that no refs need to be taken for the cached set

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11829>

ir3/print: Manual formatting fixups

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

ir3: Reformat source with clang-format

Generated using:

cd src/freedreno/ir3 && clang-format -i {**,.}/*.c {**,.}/*.h -style=file

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

freedreno: Add some options to .clang-format

In preparation for reformatting ir3.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

ir3: Manually reformat some places

clang-format does a bad job with a few tables and macros, and there were
some places it was doing wonky things because comments were longer than
80 characters and it tries to fix that without reformatting the comment
itself. Add magic comments to tell it to turn itself off and retab those
places manually (well, with a regex!).

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

ir3: Update .editorconfig and .dir-locals.el

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

ir3/lower_parallelcopy: Don't manually set wrmask

It's automatically set. This avoids some weird line wrapping with
clang-format.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

ir3: Add ir3_collect() for fixed-size collects

This avoids having the specify the size, and fixes weird formatting with
clang-format.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

ir3: Add missing include to ir3_parser.y

This prevents build errors in the generated ir3_parser.h when
clang-format reshuffles the header includes.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11801>

vulkan/wsi/wl: add wl_shm support for lavapipe.

This adds swrast support for rendering from lavapipe.

Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11819>

Revert "ci/panfrost: Temporarily disable sun50i/RK3288"

Back by popular demand etc

This reverts commit 9868b15164488f3a87d3204ff8788e9ce2f9c132.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11828>

kmsro: Add mali-dp

The NXP LS1028A SoC includes a mali-dp display controller and a vivante
GPU. This allows using the Etnaviv driver on this SoC.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11419>

st/xa: Mark default xa_get_pipe_format case unreachable.

Fix defect reported by Coverity Scan.

Uninitialized scalar variable (UNINIT)
uninit_use: Using uninitialized value fdesc. Field fdesc.format is uninitialized.

Fixes: 9f2f5b3d7fd ("st/xa: Initial import of the xa state-tracker and the xa-vmwgfx target.")
Suggested-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9035>

util/meson: include inc_gallium

There are several dependencies on headers from
   /gallium/include/pipe/

which currently mean that dependencies on util
must include gallium to compile.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11811>

i915g: Fix backface stencil when front_ccw is set.

If we're doing separate front/back stencil ops, then in the absence of a
hardware bit for setting front_ccw (which you can already see from how we
manage cull face) we need to flip front/back stencil register settings.

Closes: #4974, #4975
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11706>

i915g: Fix dumping of 3DSTATE_BACKFACE_STENCIL_OPS.

Its length was overly long and it meant we skipped the name of
3DSTATE_BACKFACE_STENCIL_MASKS.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11706>

radv: lower primitive shading rate in NIR

This allows more potential compiler optimizations if the value is a
constant or from a scalar load.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11579>

turnip: Fix allocation size for vkCmdUpdateBuffer.

tu_cs_alloc() takes a size in dwords, not bytes.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11743>

lavapipe: expose more storage-image features

I believe these Gallium caps imply these Vulkan features. If they don't,
then we got things wrong on the Zink side as well ;)

While we're at it, query shaderStorageImageWriteWithoutFormat based on
the PIPE_SHADER_CAP_MAX_SHADER_IMAGES cap. This matches what the gallium
OpenGL frontend does.

This brings Zink on Lavapipe up to OpenGL 4.5.

For some reason, a bunch of PBO tests starts failing on CI for Zink, but
that doesn't seem like a Lavapipe problem...

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10467>

llvmpipe: only report supported shader-image formats

LLVMpipe has a rather limited implementation of shader-images, so let's
limit what we report to the formats required by the spec, as all of
those seems to work fine.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10467>

lavapipe: query formats for shader-image support

Assuming all formats are supported here isn't a good plan; we
don't actually support all formats.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10467>

llvmpipe: reject unsupported shader-image formats

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10467>

ci/panfrost: Temporarily disable sun50i/RK3288

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11825>

docs: Add docs for running a local Mesa build

I'm tired of explaining this to people.  Let's put it in the actual
project docs.

Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11809>

intel/compiler: document register types

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11677>

aco: remove (wrong) GCC array-bounds warning

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

aco: refactor SDWA opcode validation

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

aco/meson: remove unnecessary dependencies

Also moves idep_vulkan_util_headers to /tests/meson.build

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

aco: add 'common/' and 'llvm/' prefix to #includes

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

aco: add missing Licenses and remove Authors from files

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

aco: reorder and cleanup #includes

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

aco/util: replace DIV_ROUND_UP(n+1,m) by n/m+1

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271>

broadcom/compiler: fix add vs. mul

Spotted by a compile warning

Fixes: 7f61ff7b4d1 ("broadcom/compiler: Merge instructions more efficiently")

Reviewed-by: Iago Torral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11764>

broadcom/compiler: use correct flag enum

They have the same value, so no functional change

Reviewed-by: Iago Torral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11764>

intel/compiler: Track latency/perf of LSC fences

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11759>

panfrost: Fix GPU ID for t76x in get_perf_config

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11824>

v3dv: expose geometry shaders

A run of dEQP-VK.*geom* has 6328 test passes and 16 failures. 14 of
these failures are related to a CTS bug affecting atomic operations
for which I submitted a CL to Khronos. Another fail looks like it might
be a case of the error threshold in CTS being slightly low. There is
only one test fail in that list that looks like it might be an actual
driver bug, but it is a variable pointers test, so it might be unrelated
to geometry shaders in the end.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

broadcom/ci: update expected fails for v3dv after enabling geometry shaders

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: remove deferred vkCmdClearAtachments path

This was required to handle the case of secondary command buffers that
did not have framebuffer information available from the primary, since
we used to have an implementation that required this information to
be available for the fallback path of vkCmdClearAttachments. Now that
we can handle our our attachment clears in the current subpass by
emitting draw calls, we no longer need the framebuffer information to
be available and we can remove this.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: remove fallback path for vkCmdClearAttachments

Now that we support layered clears this code is dead.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: implement layered attachment clears

Now that we have geometry shaders, we can leverage this to implement
clears of layered framebuffers by adding a geometry shader in our
clear pipelines that redirects the clear rect to the appropriate
layer in the framebuffer, instead of falling back to emitting separate
clear jobs for each layer, which is a lot slower.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

broadcom/compiler: don't ignore constant offset on per-vertex input loads

Fixes:
dEQP-VK.clipping.user_defined.clip_distance.vert_geom.{5,6,7,8}

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

broadcom/compiler: handle compact input arrays for geometry shaders

Clip distance arrays will come as compact array variables, so we need
to handle them as such, like we did for vertex inputs.

Fixes:
dEQP-VK.clipping.user_defined.clip_distance.vert_geom.{1,2,3,4}

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: fix copy buffer to image TFU path for 3D images

This bug was exposed by the following CTS tests (requires
geometry shaders):
dEQP-VK.geometry.layered.3d.*.readback

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: handle QUNIFORM_FB_LAYERS

There is a bit of a corner case here for secondary buffers that
don't inherit framebuffer information, since those won't have
access to the number of layers in the framebuffer by the time
we get here. Since we only emit this to sanitize the value of
gl_Layer produced from geometry shaders, I think it is probably
fine to lie about it and just use the maximum number of layers
we support. The only issue with that is that broken shaders that
emit out of bounds layer indices (which the spect states may lead
to unfined shading results) may cause undefined results.

In the future we could do better and patch the uniform streams
later when the secondary is executed inside a primary, since we
will have the required framebuffer information at that point,
but for now this seems like a reasonable compromise.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: emit state packets for geometry shaders

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

broadcom/compiler: create a helper for computing VPM config

This code is the same across drivers.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: add support for geometry shaders to pipelines

This gets our graphics pipelines (and pipeline cache) to accept
and compile geometry shader modules.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

broadcom/compiler: track if geometry shaders write gl_PointSize

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: add a few more broadcom shader stage helpers

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: define a generic helper to create binning pipeline stages

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: extend broadcom stages to include geometry

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

v3dv: always free pipeline stages after compiling

We were not freeing this when the pipeline was loaded from the cache.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11783>

crocus: add GL_CLAMP emulation in driver again.

I removed this because I thought the state tracker could handle it,
and it really should handle it, but it has some minecraft side effects
I'm unsure about. This fixes the problem for now, we can revisit it later.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11822>

crocus/gen8: fix wrap mode needs border color.

missing piece here.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11821>

agx: Flip point coordinates because OpenGL

Fixes dEQP-GLES2.functional.shaders.builtin_variable.pointcoord

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Handle point coordinates

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Identify triangle/lines vs point varyings

Seems to allow skipping over point coords?

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Unpack varying descriptors (1x)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Lower PIPE_CAPF_MAX_POINT_WIDTH to hw limit

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Set bit for psiz

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Set point magic bit in rasterizer

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

asahi: Set bits in UNK11 needed for points

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11818>

zink: fix mem info query to be more permissive

airlied was right, this was way too conservative

Fixes: ff4ba3d4a77 ("zink: support PIPE_CAP_QUERY_MEMORY_INFO")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11803>

asahi/decode: Print some IOGPU stuff

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi/decode: Print clear/store pipelines

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Garbage collect senseless cmdbuf struct

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Consolidate some magic numbers

This is a bunch of software data structures concatenated, not a true command
buffer like previously thought.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Allocate global IDs

Use the same UABI as Metal. One less hack, trying to rule out possible
differences to Metal...

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Allow specifying an encoder ID

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi/decode: Fix decoding of draw calls

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi/decode: Handle CULL packets

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi/decode: Fix up high word

Not sure what's happening here, there's some magic for sure.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Identify more unknown fields in the memmap

From validating the memory map of a Metal sample and seeing what goes wrong.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Reserve more space to stop a command buffer

Let's be safer.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi/decode: Check fewer zeroes after a command buffer

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Make track_free safer

Ensure that we don't free the same BO multiple times, which can lead to later
segfaults in decode.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi/decode: Only dump mapped allocations

This matches the hardware's view of memory and helps catch spurious reads. (One
symptom of messing this up -- besides broken rendering -- is a kernel warning
about a "bogus texture handle 0".)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Remove unused bo_access property

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Add missing copyright/guards for magic.c/h

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

asahi: Move fixed internal shaders to agx_blit.c

A more natural place to put them, as the clear/store shaders use a similar
mechanism as the reload shaders. Now magic.c is exclusively kernel side piping.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11815>

panfrost: Set register allocation in the v7 RSD

Required to hit full occupancy. Now that this information is piped
through the compiler, we can do this correctly.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>

panfrost: Generalize pan_blitter's reg count assert

Needed to avoid crashing on Bifrost/Valhall.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>

panfrost: Enable more tiler levels if we can

Boosts glmark2 scores on Mali G52.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>

panfrost: Query tiler features

We need the maximum levels to configure the hierarchy mask correctly. We
should also respect the bin size...

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>

panfrost: Zero depth_source in vertex shaders

Spurious assignment.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>

panfrost: Don't set zs_update_operation in vertex shaders

Spurious assignment.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>

panfrost: Add a performance counter dump utility

This uses Antonio's src/panfrost/perf for all the heavylifting, just
like the Perfetto producer. Unlike the Perfetto producer, it has no
dependencies and is a lot less useful. But it's a good smoke test.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733>