spirv: Move the emit a 'return value' store logic into own function

Right now, only the structured CF path emits 'return value' stores when
an SpvOpReturnValue opcode is found. Move the emit 'return value' logic
in a separate function so we can use it from the unstructured path as well.

v2 (Karol): rephrased and removed unstructured changes

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Add some debug prints

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Add a route::outside set

Acked-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Replace a tripple loop with a double loop

If there's some reason why this needs to be a tripple loop, I'm not
seeing it.  As far as I can tell, all the inner-most loop does is look
for the next remaining block not already in cur_level->blocks.  There's
no reason to re-walk the whole set every time just to do that.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Add a block_for_singular_set helper

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Sort blocks in select_fork

Hash set ordering is non-deterministic so any time we make a decision
that may affect the final structure or order of instructions, we want to
use a sorted list of blocks.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Rework some set union logic

I find the ternary a bit hard to read.  The optimization is fairly
obvious but the way it's coded makes things more dense than they
probably need to be.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Rework handling of skip targets

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Use util/list instead of exec_list

I'm trying to reduce exec_list usage in NIR.  Also, util_list has some
better helpers for a bunch of the operations this pass needs.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Clean up ralloc usage

It's really hard to track in this pass which sets are getting ralloc'd
off which other sets.  To avoid leaks, just pass a mem_ctx around
everywhere and ralloc almost everything off the one context.  We do keep
using recursion a few places where it's crystal clear what the parent
relationship is.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir/lower_goto_if: Document some data structures

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir: Add a structurizer

v2 (Karol):
  renamed pathes to paths
  use more bool
  use _mesa_set_intersects
  deduplicated some code
  fixed some typos
v3 (Karol):
  don't enable structurizer as we do this in vtn now
v4 (Jason):
  A few clean-ups due to unstructured NIR changes
v5 (Jason):
  Misc whitespace and style cleanups

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Tested-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir: Add goto_if jump instruction

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir: Add and use nir_foreach_block_unstructured helpers

These are safe to call on either structured or unstructured NIR but
don't provide the nice ordering guarantees of nir_foreach_block and
friends.  While we're here, we use them for a very small selection of
passes which are known to be safe for unstructured control-flow.  The
most important such pass is nir_dominance which is required for
structurizing.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

nir: Add a structured flag to nir_shader

v2 (Jason Ekstrand):
 - Make "structured" a property of nir_function_impl not nir_shader
 - More validation and asserts

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

spirv: rename vtn_emit_cf_list to vtn_emit_cf_list_structured

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

util/set: add _mesa_set_intersects

v2 (Jason Ekstrand): add asserts and iterate over smaller set

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>

android: panfrost: Redirect cmdstream includes through GenXML (v2)

bifrost requires both "pafrost/lib/midgard_pack.h" and "midgard_pack.h" headers

Fixes the following building error:

In file included from external/mesa/src/panfrost/bifrost/cmdline.c:33:
In file included from external/mesa/src/panfrost/bifrost/test/bit.h:31:
external/mesa/src/panfrost/lib/pan_device.h:40:10: fatal error: 'midgard_pack.h' file not found
#include <midgard_pack.h>
         ^~~~~~~~~~~~~~~~
1 error generated.

Fixes: bce1a7e9 ("android: panfrost: Redirect cmdstream includes through GenXML")
Fixes: 88dc4c21 ("panfrost: Redirect cmdstream includes through GenXML")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6322>

turnip: add missing tu_bo_list_add in CmdWriteTimestamp

Fixes an issue observed in renderdoc.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6320>

turnip: call packing functions directly for pack_gmem_clear_value

This involves rolling our own int packing functions, because the u_format
versions do clamping which differs from VK spec requirement.

This reduces the size of libvulkan_freedreno.so significantly.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6304>

turnip: rework format_to_ifmt

Make it somewhat generic instead of listing formats individually.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6304>

turnip: fix CmdBlitImage with D32_SFLOAT_S8_UINT

Fixes these dEQP tests:

dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.depth_stencil.2d_d32_sfloat_s8_uint_d32_sfloat_s8_uint.*

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6304>

turnip: delete a blit_image TODO that has already been resolved

All the blit_image tests now pass when forcing the 3D path, so this TODO
is no longer relevant.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6304>

turnip: remove dead tu_minify/typed_memcpy functions

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6304>

android: panfrost/bifrost: add libpanfrost_lib static dependency

Fixes the following building error:

external/mesa/src/panfrost/bifrost/test/bit.h:30:10: fatal error: 'panfrost/lib/midgard_pack.h' file not found
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.

Fixes: 88dc4c21e ("panfrost: Redirect cmdstream includes through GenXML")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6317>

android: panfrost: Redirect cmdstream includes through GenXML

Fixes the following building error:

external/mesa/src/panfrost/lib/decode.c:26:10: fatal error: 'midgard_pack.h' file not found
         ^~~~~~~~~~~~~~~~
1 error generated.

Fixes: 88dc4c21e ("panfrost: Redirect cmdstream includes through GenXML")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6317>

radv:  Clean up setting the surface flags.

If we don't give the init function access to the fields we will not
accidentally start setting them.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6274>

radv: When importing an image, redo the layout based on the metadata.

When importing a DMA-BUF, the image layout created in vkImageCreate may
not match the imported BO's. To make this work we redo the layout based
on the metadata of the imported image.

The original patch did a delayed allocation just as for AHB, but that
does not work for images that are not imported (but e.g. exported only).

Original patch by Simon Ser <contact@emersion.fr>

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2801
CC: mesa-stable
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6274>

anv: add no reloc flags on empty and simple bo paths.

These two paths should have no relocs.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6288>

anv: Mark anv_dump_{start,finish} as PUBLIC

In order to call these functions from gdb we need to mark both of them
as PUBLIC.

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6125>

zink: invalidate pipeline hash on more changes

some cases were missed here, causing an assert to trigger

Fixes: 1185b3f32d9 "zink: pre-hash gfx-pipeline-state"
Reviewed-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6262>

util/macros: Add ATTRIBUTE_NOINLINE definition for MSVC

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6303>

u_debug_stack_test: Fix MSVC compiling by using ATTRIBUTE_NOINLINE

Fixes: d0d14f3f ("util: Add unit test for stack backtrace caputure")
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6303>

iris: Make iris_bo_import_dmabuf take a modifier

Replace the tiling parameter with a modifier parameter. I find it more
straightforward to have this function figure out the tiling from the
modifier than to have its caller do it.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6296>

iris: Don't call SET_TILING for dmabuf imports

Calling SET_TILING on a DMA buffer with the gen12 CCS modifier can fail
unnecessarily. The main surface in the BO is Y-tiled, but the CCS portion is
linear and can have a stride that's not a multiple of 128B. Because SET_TILING
is called on the CCS plane with I915_TILING_Y, the ioctl will sometimes reject
the stride.

SET_TILING was originally used in b6d45e7f748e9ff7e198391f5ce5d1253101fedb to
fix an assertion failure in iris_resource_from_handle. Assigning the BO's
tiling_mode field is sufficient to avoid the failure.

Fixes: c19492bcdb9 ("iris: Handle importing aux-enabled surfaces on TGL")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6296>

turnip: implement VK_EXT_custom_border_color

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6093>

util/format: expose generated format packing functions through a header

Some of the generated functions can be useful without going through the
format table (filling border color struct in turnip). By not calling these
functions through the format table, we should eventually be able to garbage
collect the unused packing functions, and also allows LTOs to happen.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6093>

panfrost: add missing dependency on midgard_pack.h

Fixes build failing when libpanfrost is built before midgard_pack.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6093>

panfrost: Fix border colour

The border colour was being set to the red colour for all channels.
Fixes tex-border-1 from piglit.

Fixes: f74186baa5e ("panfrost: XMLify Midgard samplers")
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6298>

nir/lower_idiv: Port recent LLVM fixes to emit_udiv

This change fixes off-by-one results in corner cases such as
0xffffffff / 0x11111111. For details refer to LLVM bug 46212.

Fixes: 8b98d0954e6 ('nir/lower_idiv: add new llvm-based path')
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6281>

aco: do not set valid_mask for POS0 exports on GFX 10.3

This hardware issue seems only present on GFX10.

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/6278>

panfrost: XMLify enum mali_format

Note we are just handling the index part of the format. This is *not*
the full format, which would include the swizzle (or v7 equivalent) and
the sRGB flag. But in the interest of incremental progress, let's move
this part over first and save on decoding complexity.

To avoid substantial churn from prefixing FORMAT to format names, we
special case the enums to avoid the prefix. This is undesirable but
reduces churn, especially since format handling is slated for an
overhaul soon to accomodate v7

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify exception access

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify MSAA writeout mode

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify Block Format

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Drop unused mali_channel_swizzle

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify Bifrost textures

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify Midgard textures

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify Bifrost samplers

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify Midgard samplers

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Don't mask coverage mask to 4-bits

While it is correct for MSAA 4x, it will break for MSAA 8x and 16x, and
it is only dubiously correct for no-MSAA. Drop the mask.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Simplify depth/stencil/alpha

The alpha test is lowered by the frontend so we can drop that unused
code path, and stencil state can be computed at CSO create time to
reduce draw-time complexity.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Simplify zsa == NULL case

Stencil fields are only used if stenciling is enabled.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify stencil test

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify UBOs

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify viewport

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify wrap modes

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify stencil op

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify mali_func

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify draw_mode

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: XMLify job_type

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

pan/decode: Add helper to dump GPU structures

Based on generating unpack/print.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Redirect cmdstream includes through GenXML

This will provide a way to incrementally upgrade.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Build midgard_pack.h via meson

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Adopt gen_pack_header.py via v3d

We fork gen_pack_header.py from v3d, which in turn forks GenXML from
Intel. Selected changes for Mali:

* Custom prefix/vendoring
* Cull supported types
* Drop <register> and <packet> support (no cmdstream)
* Pack 32-bit words instead of 8-bit bytes
* Inline packing helpers
* Introduce exact attribute
* Introduce general minus(..) and shr(..) modifiers
* Introduce word:bit syntax
* Introduce prettyprinters

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

panfrost: Add stub midgard.xml

This will be filled in incrementally.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6195>

util: Fix up indentation in the generated format tables code.

I did this as a separate commit to make the previous one more reviewable.

Acked-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5826>

util: Change a codegenned switch statement to a nice little table.

This saves us 13 to 35kb on release drivers in my builds.

Acked-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5826>

util: Split the pack/unpack functions out of the format desc.

This gives the compiler a chance to GC pack/unpack functions separate from
the format descriptions.  For drivers that use everything, this is
+10-20kb, while for libvulkan_intel it's -1.3MB.

Bug: https://bugs.chromium.org/p/chromium/issues/detail?id=1048434
Acked-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5826>

docs: update calendar for 20.2.0-rc2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6292>

docs: update calendar for 20.2.0-rc1

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6292>

radeonsi: fix compute-based culling with VERTEX_COUNTER_GDS_MODE == 1

Discovered when testing Sienna Cichlid.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6282>

radeonsi: disable NGG culling on gfx10.3 because of hangs

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6282>

zink: handle more draw modes

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6269>

zink: change pipeline hashes to index based on vk primitive type

this is a bit more convenient since we always support vk types but not
necessarily gallium types

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6269>

anv/nir: Unify inputs_read/outputs_written between geometry stages

inputs_read/outputs_written are used for a shader stage to
determine the layout of input and output storage. Adjacent stages must
agree on the layout, so adjacent input/output bitfields must match.

Most of the time, cross-stage optimizations make that happen anyway,
but there are some cases (with special values like clip distances and
point size) where this doesn't happen.

Fixes crashes in dEQP-VK.subgroups.*.framebuffer.*_tess_eval

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3210
Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6102>

panfrost: Implement panfrost_query_dmabuf_modifiers

v2: Only advertise AFBC if PAN_MESA_DEBUG=afbc is set. This is to avoid
sharing AFBC buffers as SCANOUT until corresponding kernel bugfixes are
landed. Technically a kernel issue but let's prevent a regression.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me> [v1]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Ensure AFBC slices are aligned

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Allocate enough space for tiled formats

We need to align and possible reserve space for a header.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Enable YTR where allowed

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Implement YTR availability check

Depends on format.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Choose AFBC when available

There are lots of reasons we might fallback on u-interleaved tiling, but
when we can use AFBC, it's a big win.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Import staging routines from freedreno

For software access to AFBC textures.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Respect modifiers in resource management

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Account for modifiers when creating BO

If specified, use the modifier directly. If unspecified, pick our own
modifier.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Use modifier instead of layout throughout

Instead of converting back and forth we should stick with fourcc codes
as the canonical layout definition. Furthermore modifiers allow all the
variants of AFBC to be encoded canonically, whereas the previous enum
does not (info about YTR is encoded out of band, for instance).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Introduce create_with_modifier helper

As a stepping stone to full modifier support.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Remove hint-based AFBC heuristic

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

panfrost: Set `initialized` more conservatively

In case Gallium doesn't ask for direct but we give a direct mapping
anyway.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

gallium/dri2: Support Arm modifiers

This is needed to share both Utgard-style tiled and a subset of Arm
FrameBuffer Compression (AFBC) tiled framebuffers across processes in
Wayland.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Tested-by: Icecream95 <ixn@keemail.me>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6159>

zink: print error when getprocaddr fails for extension functions

make this more visible on errors

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6271>

radv: Update CI expectations for the recent descriptor indexing regressions.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6285>

iris: Add support for MESA_SHADER_KERNEL in the disk cache

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

iris: Upload kernel inputs with system values

Clover doesn't upload a cbuf0 but instead provides the kernel inputs as
part of the pipe_grid.  The most obvious thing to do is to upload them
along with system values.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

iris: Copy dest size from the original intrinsic in setup_uniforms

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

iris/disk_cache: Stop assuming stage == cache_id

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

iris: Add a kernel_input_size field for compiled shaders

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

iris: Use blob_write_uint32 for num_system_values

We read it with blob_read_uint32; we should write it as uint32 as well.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

intel/compiler: Allow MESA_SHADER_KERNEL

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>

intel/compiler: Use C99 array initializers for prog_data/key sizes

This is way better than a pile of STATIC_ASSERTs.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6280>