zink: remove refs from buffer rebinds

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11692>

zink: move resource object ref to batch in init_storage_object

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11692>

zink: move resource object ref to batch in invalidate hook

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11692>

zink: add util function for transferring resource refs to batch

there's no point in doing unref->ref if it's known that this will happen

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11692>

relnotes: more lavapipe features

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11738>

features: more lavapipe extensions

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11738>

lavapipe: implement KHR_separate_depth_stencil_layouts

this is a conformant implementation

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11738>

lavapipe: implement EXT_separate_stencil_usage

this is a conformant implementation

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11738>

pan/decode: Handle IDVS jobs on Bifrost

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11413>

panfrost: Add secondary shader XML fields

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11413>

panfrost/genxml: Decode Bifrost index-driven vertex jobs

Bifrost has a new type of job called 'index-driven vertex', which takes
in two shaders. The primary shader, called for all values, performs
position shading to a cacheline-aligned attribute buffer. The secondary
shader, called for values still live after culling, performs varying
shading.

It is the responsiblilty of the implementation to ensure that vertices
from different instances do not share a cacheline.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11413>

iris: Use isl_surf_get_uncompressed_surf

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11647>

iris: Don't leak the surface if uncompressed re-interp fails

Fixes: a032a9665f27 "iris: Enable PIPE_CAP_SURFACE_REINTERPRET_BLOCKS"
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11647>

intel/isl: Add more cases to isl_surf_get_uncompressed_surf

We can actually create array surfaces instead of requiring single-slice
in a few cases.  This does require us to be very careful about our
checks, though.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11647>

intel/blorp: Use isl_surf_get_uncompressed_surf

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11647>

intel/blorp: Adjust the compressed copy rectangle before convert_to_single_slice

It doesn't matter for the actual copy rectangle and this makes the
asserts a bit nicer as we don't need to bother with the intratile
offsets because there aren't any yet.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11647>

intel/isl: Pull the uncompressed surface view code from anv

This adds a helper isl_surf_get_uncompressed_surf for creating a surface
which provides an uncompressed view into a compressed surface.  The code
is basically a direct port of the uncompressed surface code from the
Vulkan driver which, in turn, was a port from BLORP.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11647>

crocus: Reduce SSBO alignment requirements from 64B to 4B.

Mirroring a change in iris; see the previous commit for rationale.
One difference is that crocus does sometimes use CPU maps on Atoms
for *reads*...but writes don't go through CPU caches.  Should be OK.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11727>

iris: Reduce SSBO alignment requirements from 64B to 4B

Back when SSBOs were first enabled in i965, we tried to work around
issues where the CPU and GPU were incoherently writing to the same
cacheline by forcing an alignment such that different sections of
data would fall in different cachelines.  This seems wrong.

On integrated GPUs with LLC, CPU and GPU writes should be coherent.
On integrated GPUs without LLC, we either enable snooping (so they
are again coherent), or we use WC maps (so the CPU cache isn't used).
Discrete GPUs always use WC maps (so the CPU cache isn't used).

This should work.  In other words, I think the increased alignment
was just working around coherency problems on atoms that have been
fixed in the intervening 6 year time period.

Untyped surface messages require 4B alignment.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5016
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11727>

panfrost: Eliminate redundant flushes with AFBC

We have all the information we need easily available.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11077>

panfrost: Split "flush writer" from "flush accessing"

Since writers are counted in the accessor set, the old implementation
had a redundant flush in the "flush accessing" case. To fix, split the
two modes into separate functions, removing the redundancy and offering
a much more ergonomic API.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11077>

panfrost: Do tracking of resources, not BOs

Squashed together with commits from Boris's original dependency tracking
cleanup series.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11077>

panfrost: Wrap occlusion query in pipe_resource

For proper tracking.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11077>

panfrost: Express dependencies as resources, not BOs

This is a more ergnomic API, and will allow further optimizations.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11077>

panfrost: Replace the batch->bos hashmap by a sparse array

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11077>

nir: Free the instructions in a DCE instr removal.

No significant change in shader-db time (n=11), but should be a little win
for memory usage by the compiler.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11628>

nir: Use remove_and_dce for nir_shader_lower_instructions().

Reduces the work that other shader passes have to do to look at dead code,
and possibly extra rounds around the optimization loop if dce wasn't the
last pass in it.

shader-db runtime -1.12919% +/- 0.264337% (n=49) on SKL.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11628>

nir: Add a nir_instr_remove that recursively removes dead code.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11628>

anv: Fix unused var warning on release builds from an assertion.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11628>

zinK: tweak shader module update -> pipeline combined_dirty conditional

make this more explicit

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11691>

zink: remove shader_id

now that shaders are per-program, shader_id is no longer useful

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11691>

zink: remove gfx program slot mapping

if shaders aren't being shared between programs, they can instead directly
assign io between the shader stages such that only the producer and consumer
pair need to have their slots match up

this requires that each gfx program struct stores base nir for a given shader
in order to avoid breaking io for other programs which reuse shader states

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11691>

zink: set gfx program shaders and generate internal tcs during program creation

the base shader states will never change for a program, and tcs will never
change its io, so there's no point in having this in the shader update hotpath

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11691>

zink: split up shader cache per-stage

this reduces overhead from lookups by ensuring that the number of entries
in a given table are minimized

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11691>

zink: make shader cache local to a single program

sharing caches makes things complicated and doesn't yield much benefit

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11691>

gbm: Load backend based on DRM device driver name

After attempting to use the GBM backend specified by the user
via an environment variable, if any, but before falling back
to the built-in GBM backends, attempt to dlopen
libg<DRM driver name>_gbm.so in the GBM backend library search
path (Defaults to "$libdir/gbm") and initialize a device using
it. This enables automatic backend discovery for devices that
do not provide a DRI driver.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Support dynamically loading named backends

If the user specifies a backend name explicitly
via an environment variable and it is not in the
list of built-in backends, attempt to load it at
runtime.

runtime-loaded backends get a new gbm_backend_desc
struct instance for each device using them (A
small increase in memory usage to eliminate the
need for the locking and bookkeeping sharing them
would require), so these structures need to be
freed when destroying devices using runtime-loaded
backends.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Put common device creation in a helper function

Subsequent changes will use this same logic to
instantiate devices with the correct version and
set up their backend reference.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Rename the DRI backend from gbm_dri.so to dri

A subsequent change will allow loading backends
from DSOs specified by the GBM_BACKEND environment
variable. The exact DSO name and path are derived
by the common loader code and will be of the form:

 <gbm_backend_path>/<GBM_BACKEND>_gbm.so

E.g., a user would set the environment variable to
"external" to load "external_gbm.so". Users will
also still be able to explicitly request any
builtin backends by name as well, so this change
helps keep the environment variable syntax
consistent between internal and external backends.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

meson: Add a GBM backends search path build option

This is the list of locations where GBM will look
for alternate backend implementations based on the
DRM driver name. It defaults to $libdir/gbm.

On Android, this path is currently hard-coded to
/vendor/lib[64]/gbm.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

loader: Factor out driver library loading code

Separate the path-building and dlopen() portion of
loader_open_driver() from the DRI extension
loading logic. The former will be shared by the
GBM backend loading logic in a subsequent change.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Rename backend description list to builtin_backends

This just makes it clearer the array isn't a global
list of all backends in use.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Add backend ABI-check test

This is based on the wayland EGL library ABI test
framework. The helper macros were copied from
there and expanded to support more than one
struct/type, and to additionally check member
type compatibility by default.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Version the GBM backend interface

Define a version number for the interface GBM uses
to offload work to its backends/drivers. Store the
version in the backend interface structs provided
to the loader by backends, as well as in the core
interface struct provided to backends by the GBM
loader code to backends.

The backend can create structures of any version
it supports, which can be greater or less than the
interface version specified by GBM in the core
interface structure. Hence, GBM will need to take
care to check the backend version before accessing
any members added to structs defined in
gbm_backend_abi.h after this change.

Similarly, the backend may need to check the
interface version supported by the GBM library
before passing back data in any structure members
that require the GBM library to interact with
them for correct operation. For example, if for
some reason a structure defined in
gbm_backend_abi.h gained a field which was a
pointer to memory allocated by the backend and
freed by GBM, the backend should avoid allocating
this memory if the GBM library did not specify an
interface version new enough to indicate that it
was aware of the new structure member.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Move majority of gbmint.h to gbm_backend_abi.h

The content of gbm_backend_abi.h are nearly an
identical copy of the old gbmint.h. Only minimal
modifications have been made to ease history
tracking. Subsequent changes will further
formalize the ABI structures in gbm_backend_abi.h

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Add gbm_core struct to export code to backends

The GBM core/loader code defines one helper
function used by both itself and the built-in DRI
backend. Presumably, external backend authors
would want to use such functions as well, so
package them into a single struct that will be
passed explicitly to externally loaded backends in
subsequent changes.

Another option considered was to simply export
the gbm_format_canonicalize() function directly,
optionally renaming it to better indicate it is
intended only for "internal" use first. However,
even with a rename, this would expose it to
potential use by applications as well, which is
not ideal, as it is not intended to be part of
the application-facing GBM ABI.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Give gbm_device a reference to its backend

This will be used for looking up the backend
again at destruction time to perform any backend-
agnostic cleanup.  To facilitate that, also
dispatch device destruction to the backend manager
code.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Give getenv backend override its own function

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Consolidate env var and default backend loops

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Create device directly in find_backend

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Inline load_backend function content

It was a functional no-op with the currently
available backends (Only DRI), and its implied
intended functionality is implemented elsewhere
in subsequent patches.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

gbm: Remove stat and refcount fields from gbm_device

These were unused.

Signed-off-by: James Jones <jajones@nvidia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9902>

util/vbuf: check 3-component 16bit int formats for translation

some drivers don't support these

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11716>

lavapipe: handle null vertex buffers more gracefully

this should propagate down to be handled correctly in llvmpipe

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11723>

panfrost: Only clear existing color buffers

    in case that only one cbuf allocated but all clear bits set,
    the driver will crash due to null pointer dereference.

Signed-off-by: luc <onion0709@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11515>

Revert "panfrost: Fix crc_valid condition"

This "fix" is no longer needed now we write CRC data correctly for
partial writes.

This reverts commit a78487f795e81c8fc27c2d1054f58dd35eda8c3a.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11446>

panfrost: Set bound dimensions to framebuffer size

Otherwise CRC calculation will miss the pixels outside the extent
bounds but still in a tile that is rendered.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11446>

ac: fix win32 build

Fixes: e2e9dd44f422 ("ac/surface: Handle non-retiled displayable DCC correctly for modifiers.")
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11725>

asahi: Use XML for interpolation packet

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Enable primitive restart

Passes dEQP-GLES3.functional.primitive_restart.* except for a flake.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Set data_valid for the depth buffer

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Identify attachment length field

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Use GenXML for main bind fragment

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Extend IOGPU header to contain encoder

Let's squash together all the resulting zeroes. Trying to discern some structure out of this.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Move IOGPU header to XML

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Identify "command buffer" size field in map

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Parametrize software "command buffer" size

Will vary depending on # of attachments.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Sync attachment magic with asahi demo

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Add XML for the attachment structure

We need a lot more control over these magic structures to get Z/S attachments working correctly. This is an early start.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Skip over holes in the vbufs

Fixes dEQP-GLES2.functional.vertex_arrays.multiple_attributes.stride.3_float2_17_float2_17_float2_0

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Comment on an embedded data structure

This file makes me sad :'(

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Flush for accesses to Z/S buffer

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Implement the stencil test

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Prepack rasterizer faces

A bit more efficient and will allow easy implementation of the stencil test.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

asahi: Fix meson dependency on packing in compiler

Closes: #5030
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Fixes: 6a657b17b9b ("agx: Remap fragment shader varyings explicitly")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11730>

radv: implement VK_EXT_multi_draw

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: add a gfx10 bug workaround for NOT_EOP

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: split indexed draw cases based on whether drawid is used

fewer conditionals in the draw loop = less overhead

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: determine if hardware can emit NOT_EOP before emitting

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: emit drawid for multidraws

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: emit NOT_EOP for multi indexed draws

this is never emitted currently because drawCount is always 1

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: use multidraw iteration for direct draws

there's still no arrays of draw info being passed, but this takes care of
all the work for it

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11532>

radv: use null vertex descriptor if num_records=0

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11642>

radv: adjust num_records when offset>stride

If an attribute's offset is larger than the stride, the compiler will
increase the vertex index and use offset%stride instead as the offset.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5007
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11642>

radv: don't ever convert num_records to bytes if it's zero

If num_records==0, this conversion doesn't work. We should just keep it at
zero.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5007
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11642>

crocus: enable threaded context support

Reviewed-by: Zoltán Böszörményi <zboszor@pr.hu>
Tested-by: Matti Hämäläinen <ccr@tnsp.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11710>

crocus: add unsync transfer pool

Reviewed-by: Zoltán Böszörményi <zboszor@pr.hu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11710>

crocus/tc: init/deinit threaded resource

Reviewed-by: Zoltán Böszörményi <zboszor@pr.hu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11710>

crocus: use threaded context base classes

should be a pretty non-functional change

Reviewed-by: Zoltán Böszörményi <zboszor@pr.hu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11710>

crocus: support rebinding streamout target buffers

Reviewed-by: Zoltán Böszörményi <zboszor@pr.hu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11710>

ac/surface: Handle non-retiled displayable DCC correctly for modifiers.

There is some hardware with num_render_backends == 1, but the number
of render backends in GB_ADDR_CFG > 1. Turns out this can be turned
off by making them rb unaligned which is valid with only 1 render
backend.

Fixes: 0833dd7d124 ("amd/common: Add support for modifiers.")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10575>

features: mark off EXT_vertex_input_dynamic_state for lavapipe

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11724>

features: add VK_EXT_multi_draw

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11722>

features: fix ARB_shader_group_vote -> GL_ARB_shader_group_vote

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11721>

features: mark off some zink features

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11720>

features: mark off line rasterization for lavapipe

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11719>

relnotes: add some line feature updates for lavapipe

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11715>

asahi: Handle nonzero first_level

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11718>

asahi: Wire in u_blitter

Fixes dEQP-GLES3.functional.fbo.color.repeated_clear.blit.tex2d.rgba8

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11718>

asahi: Track more Gallium state

Needed to feed u_blitter.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11718>