docs: inline contents.rst into index.rst

The idea of having a single file containing the ToC is not really how
things are done in Sphinx, and kinda makes it harder to structure
documentation more naturally. This was just something I did to mirror
what we used to do for the old HTML-only version of the docs, to ease
the transition and to de-clutter index.rst.

Now that the transition is far behind us, and index.rst is much cleaner,
we can finally start inlining this.

In the long run, I expect most of these to be moved to separate "chapter
articles" that summarize what these topics are, and thus disappear from
here.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7709>

docs: mention egl in api-list

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7709>

docs: do not explicitly call out es-versions

There's no reason to explicitly call out GL ES versions here, as these
are all major versions there are, and it's very unlikely that more
versions will emerge.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7709>

docs: use external link-references

The rST code here is much more to the point and easy to read if we
define the links as external link-references instead of inlining them.

This will make the next few patches much easier to grok.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7709>

docs: move major versions history out of front-page

The features added in each major version is also unlikely the first things
someone wants to know about Mesa. So let's move this into the
versions.rst article.

This documentation is severely out of date anyway, and as it doesn't
seem like anyone is interested in documenting this any more, we should
probably consider driopping versions.rst entirely in the longer run.

But for now, this makes the front-page much more approachable.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7709>

docs: break project history out of front-page

The project history isn't usually what a user wants to read first when
they read about a project, so let's remove it from the front-page of the
documentation.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7709>

zink: revert to old load_ubo implementation

It turns out, the load_ubo to load_ubo_vec4 implementation isn't quite
enough for us, for a few reasons:

1. We use a single array of uvec4s for our UBOs, and to handle 64-bit
   values in UBOs, we need further lowering.
2. The whole vec4 stuff seems a bit hard to reconsile with glsl 4.3
   packing as well as PackedDriverUniformStorage.

In addition to the above, this fixes several piglit tests that *aren't*
part of quick_gl, which is what I've been running. So this doesn't even
work correctly right now.

So let's go back to what we had before instead.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3643
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7858>

zink: add alternate ubo loader in ntv

This is a loader that handles more cases than the current vec4-based
loader, that is a bit easier to extend to more features.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7858>

util/slab: do not dereference NULL-pointer

This used to not be a problem, because these mutexes were the first
members of this array, meaning that we ended up trying to lock/unlock
NULL mutexes. But this isn't guaranteed to be allowed, so we were
relying on luck here.

Recently, this changed. We introduced asserts for NULL-pointers, and
changed the behavior in a way that leads to crashes in release-builds.
This means we can't rely on luck any longer.

Fixes: e3171037539 ("c11/threads: Remove Win32 null checks")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3903
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7853>

radeon/vcn: hevc main10 profile decoding pitch fix

for hevc main10 profile decoding, dpb buffer alignment fix

Signed-off-by: Ruijing Dong <Ruijing.Dong@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7827>

d3d12: Fix GCC warnings for missing function prototypes

Most of these are adding 'static', for functions that are local
to a translation unit but weren't declared static.

There's one instance of a missing include for bringing the prototype
into the translation unit, one function missing a return type (default-int),
and one which added inline to avoid it being considered unused in some sources.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

microsoft/compiler: Add missing 'return' to switch case

Fixes: b9c61379 ("microsoft/compiler: translate nir to dxil")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Fix use of incorrect clear color variable

Fixes: 2ea15cd6 ("d3d12: introduce d3d12 gallium driver")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

microsoft/compiler: Misc fixes caught by GCC

* Fix const-correctness on dxil_mdnode pointer arrays
* Fix warning for a missing scope in a case block

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Misc fixes caught by GCC warnings / code inspection

* A const-correctness issue
* Some uninitialized variables
* A missing copyright header
* A comment with a /* in it when only * was needed

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

microsoft/compiler: Fix unhandled switch case warnings

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Fix unhandled switch case warnings

Some D3D enums have invalid values (e.g. NONE) in the enum,
so default: makes sense to catch invalid values in those cases.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Clean up d3d12_compiler.h

Move some stuff from d3d12_context.h to d3d12_compiler.h, and
fix d3d12_compiler.h to not include d3d12_context.h.

This serves two purposes:
1. Putting declarations and definitions where they really belong.
2. Making it so only C++ code needs d3d12.h simplifies the helpers
   we need to add to add to support d3d12.h for Linux.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Remove Windows-specific macros

These all have trivial standard or Mesa replacements

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Fix signed-unsigned comparison warnings

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Fix brace-initialization issues

For C++ code, we shouldn't use {0}, instead use {} to default-
initialize everything. For code with nested structs/unions, make
sure there's enough braces to appropriately scope initialization.

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

microsoft/compiler: Remove dead code/variables

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

d3d12: Delete unused local variables

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7780>

gallium/tests: fix unused-but-set-variable warning

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7713>

softpipe: fix maybe-uninitialized warning

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7713>

wgl: fix maybe-uninitialized warning

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7713>

glsl: fix redefinition warning on win32

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7713>

glapi: fix unused-function warning

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7713>

util: fix mingw format-extra-args warning

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7713>

nir: don't sink instructions into loops

Repeatedly loading constants or evaluating ALU operations
in loops doesn't seem beneficial. This might increase the register
pressure, but the tradeoff seems worth it.

Totals from 13629 (9.77% of 139517) affected shaders (RAVEN):
SGPRs: 1179481 -> 1184697 (+0.44%); split: -0.03%, +0.47%
VGPRs: 978776 -> 978732 (-0.00%); split: -0.02%, +0.02%
SpillSGPRs: 51036 -> 50943 (-0.18%); split: -1.35%, +1.17%
CodeSize: 113775020 -> 113428812 (-0.30%); split: -0.34%, +0.04%
MaxWaves: 49877 -> 49881 (+0.01%); split: +0.02%, -0.01%
Instrs: 22295979 -> 22204936 (-0.41%); split: -0.42%, +0.02%
Cycles: 1637198832 -> 1626916048 (-0.63%); split: -0.64%, +0.01%
VMEM: 2403434 -> 2507645 (+4.34%); split: +4.76%, -0.42%
SMEM: 849676 -> 834576 (-1.78%); split: +0.60%, -2.38%
VClause: 412396 -> 398139 (-3.46%); split: -3.46%, +0.01%
SClause: 810480 -> 817349 (+0.85%); split: -0.19%, +1.04%
Copies: 2188260 -> 2166716 (-0.98%); split: -1.18%, +0.19%
Branches: 761204 -> 760475 (-0.10%); split: -0.15%, +0.05%
PreSGPRs: 972892 -> 981054 (+0.84%); split: -0.05%, +0.89%
PreVGPRs: 925390 -> 925420 (+0.00%); split: -0.02%, +0.02%

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7694>

d3d12: do not inspect NULL samplers

We can risk having NULL samplers in this range, so let's ignore them
while we're populating the shader-key.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7856>

aco: Fix NGG GS assert failure from the WG scan.

There was a temp which was defined in a branch but used outside,
without a phi.

Fixes: 62b5012ec35358a7c109746bfe1da33ca1b04f24
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7817>

intel/tools/aubinator_error_decode: allow 0 arguments

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7833>

intel/tools/aubinator_error_decode: allow "-" as an input file

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7833>

intel/tools/aubinator_error_decode: exit with an error on unknown option

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7833>

d3d12: lower bitfield_extract to shifts

This wasn't implemented yet, because we hadn't encountered it yet. But
now it seems we can trigger this, thanks to the nv_copy_depth_to_color
piglit tests.

This makes the test go from crash to fail, which isn't perfect, but it's
better than nothing.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7855>

Revert "radv/llvm,aco: always split typed vertex buffer loads on GFX6 and GFX10+"

It introduces regressions.

This reverts commit 6fb4babfe98e222e648eb2ab210136b7540bf039.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7852>

ci: disable check-commits

Otherwise the CI sanity check fails if the commit title exceed
80 characters

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7852>

etnaviv/drm: add some locking asserts

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7832>

etnaviv/drm: convert to simple_mtx

We do not need a full blown pthread mutex.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7832>

etnaviv/drm: fix evil-twin etna_drm_table_lock

Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7832>

v3dv: don't log out of pool memory errors for internal driver pools

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7848>

v3dv: move error string definition to debug path

No point in computing this if debug is disabled.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7848>

mesa: replace ParameterValueOffset[i] with Parameters[i].ValueOffset

to merge both arrays and remove malloc because both arrays have
the same size.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

compiler: decrease STATE_LENGTH from 5 to 4

The rework of matrix state vars made the last element unused.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

st/mesa: fix uninitialized/random clip plane state vars in lower_ucp

Fixes: 584f27326c1 - st/mesa: factor ucp-lowering logic into helper

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: allocate the attribute stack on demand

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: remove gl_texture_object references from glPush/PopAttrib stack

I don't see why this is needed and it's not used anywhere.
As long as apps don't call glDeleteTextures, nothing will release them.
And even if they do, we don't use the saved textures anywhere.

Also, BindTexture will fail for deleted textures anyway, so they can't be
popped. The existing code already binds the Name that was saved, not
the texture object that was saved.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: more optimizations in glPopAttrib (colormask, drawbuffers, coord replace)

Acked-by: Adam Jackson <ajax@redhat.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: optimize out no-op calls in glPopAttrib

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: skip _mesa_set_enable in glPopAttrib if there are no changes

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: reduce the size of gl_texture_attrib_node::Texture by about 90%

Let's get rid of the big memcpy.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: optimize saving/restoring bound textures for glPush/PopAttrib

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: reorganize gl_texture and sampler structures for glPush/PopAttrib

Put the fields saved by glPush/PopAttrib into the sub-structure declared
as Attrib. This will make glPush/PopAttrib much faster because it will
only save and restore that structure.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: add a fast path for restoring light attributes in glPopAttrib

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: add a fast path for restoring fixed-func tex state in glPopAttrib

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: rewrite glPushAttrib/glPopAttrib to get rid of malloc

This was inevitable. This gets rid of the malloc/free calls and chaining
each push group in a linked list. Now the whole attribute stack is declared
statically in gl_context.

Yes, this improves performance a lot.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: treat glPopMatrix as a no-op state change if it doesn't change the matrix

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: memset matrices at initialization to enable memcpy on it

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: canonicalize matrix in glPushMatrix to make glPopMatrix possibly a no-op

If there are no changes to the matrix, we can move pop the stack but not
update any state constants.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: consider glPushMatrix a no-op change from the driver perspective

It doesn't change any states.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: skip glMultMatrix if the matrix is identity

This happens a lot with viewperf and it causes unnecessary constant buffer
updates.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: fix crashes in the no_error case of invalid glUniform calls

I copied error conditions from validate_uniform_parameters.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

st/mesa: replace st_context::state::constants with a mask

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

st/mesa: add a faster path for uploading state parameters into constant buffers

The old path copies state parameters into the parameter list, and then
the driver copies them into a buffer.

The optional new path loads state parameters into a buffer directly.
This increases performance by 5% in one subtest of viewperf.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

gallium: add PIPE_CAP_PREFER_REAL_BUFFER_IN_CONSTBUF0

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: add helpers for drivers to load state parameters into buffers

Drivers should upload only UniformBytes of uniforms and constants,
and then use _mesa_upload_state_parameters to upload state parameters.
This allows removing one copy of state parameters.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: merge light state parameters for faster uploads (disabled)

Disabled because of CI failures.

Invoke fetch_state only once for all lights in the best case instead
of 8*8 times.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: merge matrix state parameters for faster uploads (disabled)

Disabled because of CI failures.

Instead of separate state vars for each row and invoking fetch_state 4x:
  state.matrix.modelview.row[0]
  state.matrix.modelview.row[1]
  state.matrix.modelview.row[2]
  state.matrix.modelview.row[3]

The rows are now merged and fetch_state is invoked once:
  state.matrix.modelview.row[0..3]

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: take advantage of sorted parameters in _mesa_load_state_parameters

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: put constants before state vars for ARB programs

This moves state vars to the end of the parameter list, so that state vars
can be loaded directly into a buffer instead of loaded into the parameter list.
Also, state vars don't need to be searched in the parameter list anymore,
because we will know their index range. (this will make gallium faster)

This commit just wraps a for loop around the existing code.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: put constants before state vars for ffvp

This moves state vars to the end of the parameter list, so that state vars
can be loaded directly into a buffer instead of loaded into the parameter list.
Also, state vars don't need to be searched in the parameter list anymore,
because we will know their index range. (this will make gallium faster)

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: restructure gl_light vars to match the layout of gl_LightSource uniforms

this will make uploads faster

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: optimize setting gl_Light state parameters

The order of enums is a preparation for a future commit, but if you are
guessing that I just want to copy all light params into a constant buffer
with one memcpy, you are right.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: demystify material_attrib()

the next commit would break this magic math

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: allow multi-slot program parameters

also the removed comment was incorrect.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: fix printing state parameters

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: remove redundant _math_matrix_analyse calls in fetch_state

Only program matrices needed this. Other matrices are updated in
_mesa_update_state_locked.

Acked-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: rework matrix statevar enums to remove excessive branching in fetch_state

Instead of having $matrix and $modifier as separate enums, combine them
to 1 enum $matrix_$modifier.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: don't allocate matrices with malloc

There is no reason for it. This removes a pointer indirection.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

util: add a common ALIGN16 macro for m_matrix and u_threaded_context

to prevent conflicts in the next commit

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: replace _mesa_problem with unreachable in fetch_state

let's get this out of release builds

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: don't read from destination memory when computing state parameter values

The destination memory can be uncached, e.g. a buffer object.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: skip redundant uniform updates for glUniformHandle

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: skip redundant uniform updates for glUniformMatrix

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: skip redundant uniform updates for glUniform

Viewperf does a lot of redundant uniform updates - 60-80% in some tests.
Those are sometimes the only state changes between draw calls.
This improves performance by 33% in one viewperf subtest.

If you are worried about CPU overhead in the non-redundant case,
glthread is the solution.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: move sampler condition for flushing into mesa_flush_vertices_for_uniforms

This is just a code consolidation. It might fix something, not sure.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: call FLUSH_VERTICES before changing sampler uniforms

Fixes: 9545139ce5f "mesa: skip FLUSH_VERTICES() if no samplers were changed"

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: don't print GL errors in release builds if MESA_DEBUG=silent

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: properly disallow param list reallocation

This can more easily detect shader variants adding too many state
parameters and causing a reallocation.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: track ParameterValues size separately

This is needed for multi-slot parameters.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

mesa: don't duplicate allocation code in _mesa_new_parameter_list_sized

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6946>

radeonsi/gfx10: fix overflow and primitive queries

This aligns the offsets to match the memory layout of the query buffer
defined by gfx10_sh_query_buffer_mem and calls si_launch_grid_internal
to flush caches and wait for completion of shaders prior to retrieving
results.

Acked-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/7181>

v3dv: ignore filter in TFU blit path

The filter is only relevant to handle blits that invole scaling, but
our TFU path doesn't handle any kind of scaling so we can safely
ignore it.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7845>

v3dv: add a helper to choose a compatible TFU format

We are now doing this in 3 different places so it makes sense.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7845>

v3dv: handle Z mirroring in the TFU blit path

The TFU path only activates for blits that are really copies
(no linear filtering, no scaling, same pixel format, etc.), and
we do it slice by slice, so we can easily handle mirroring of the
Z coordinate for 3D images by reversing the order of the layers
as we copy them.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7845>

v3dv: expand the formats that can be handled in the TFU blit path

Same as with other TFU paths, we only handle exact copies without
conversion, so we can rewrite the format to use a compatible TFU format
based on its texel size, which allows us to use this path with more
formats.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7845>

aco: Avoid extra bitfield padding

MSVC may insert padding for data type switches.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7785>

aco: Stub sections that don't have _WIN32 support

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7785>

aco: Fix warnings for bools in bitwise logic

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7785>