docs/systems: Update Panfrost link

There's not much on the standalone site anymore (it dates back from
before Panfrost was upstreamed), let's just keep everything in Mesa.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8148>

docs: Update Panfrost in the source tree

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8148>

docs/panfrost: Fix comment about Lima

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Suggested-by: Vasily Khoruzhick <anarsoul@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8148>

vc4: Fix typos.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8120>

radv: remove useless push constants data when resolving ds attachments

Depth/stencil resolves are only allowed inside a subpass, which means
the offset is always 0 and the draw/dispatch covers the whole image.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8127>

spirv: Remove more dead variables

SPIR-V modules can have multiple shaders (including of the same
stage), but the global variables are all declared for the whole
module.  This can result in variables with same Binding but
incompatible types, so those need to be removed before we use.

Previously, a similar issue but with a narrower scope was fixed by
6775665e5ee ("spirv: Eliminate dead input/output variables after
translation.").

This patch depends on the previous patch that prevents variables used
only in pointer initializers to be considered dead.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3686
Fixes: 3a266a18 ("nir/spirv: Add support for declaring variables")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8133>

nir: Consider pointer initializers in nir_remove_dead_variables

Between the creation of a shader (from GLSL or SPIRV frontends) and
nir_lower_variable_initializers is called, variables may refer to
other variables for initialization.  Those referred variables need to
be kept alive, so consider that in the pass.

Fixes: 7acc81056f7 ("compiler/nir: Add support for variable initialization from a pointer")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8133>

nir: Remove unused parameter in remove_dead_var_writes

And destroy the live set earlier in nir_remove_dead_variables.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8133>

nir: Fix outdated name in comment

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8133>

spirv: Implement OpArrayLength for OpenGL

Uses same NIR intrinsic as glsl_to_nir.  Make it an option so it is
easy later to move Vulkan drivers incrementally to use it.

Fixes piglit test spec/arb_gl_spirv/execution/ssbo/unsized-array-length.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3691
Fixes: 15e43907 ("iris: Enable ARB_gl_spirv and ARB_spirv_extensions")
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8136>

zink: fix gl_SampleMaskIn handling

this is an array[1], so we need to do some extra legwork here even though
probably it'd work fine even without it

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7490>

zink: fix rp hash table

hash table keys for inserted items have to be valid memory ranges for the
lifetime of the corresponding entry, so using a stack-allocated key like this
is broken and doesn't accurately return the correct renderpass

Fixes a872f4636924: ("zink: cache render-passes")

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8011>

zink: move rp hash functions further up in file

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8011>

zink: handle nir_op_find_lsb and nir_op_ifind_msb in ntv

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8153>

zink: handle nir_op_ibitfield_extract: in ntv

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8153>

zink: add nir_op_bit_count to ntv

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8153>

zink: force stencil format for stencil-only samplers and swizzle the right component

packed buffers will still return the full format when we're using only the
stencil aspect, so we need to explicitly set the format here

also due to 7ca72f172678116d29d254b786a9422b864aef3d we can't use the provided
swizzle for this case and have to force y -> x component swizzle

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7491>

mesa/st: set lower_point_size for tes/gs during program update

these may not reach the other callsite in st_precompile_shader_variant(),
so we need to ensure that we get psiz if we need it

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5885>

mesa/st: flag ST_NEW_CONSTANTS upon running nir_lower_point_size_mov

this reduces to a load_ubo after optimization, so we need to ensure that
the constant data is put in a buffer instead of relying on it happening
coincidentally

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5885>

mesa/st: handle running nir lower passes for ucp and psiz in tess stage

if tess is the last shader program to run then we need to perform these
lowering passes at that time rather than during the vs

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5885>

mesa/st: tabs -> spaces in st_program

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5885>

mesa/st: do not run lower_psiz_mov on vertex shader if geometry shader is present

this only needs to be changed on the last stage between the two

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5885>

mesa/st: run nir_lower_point_size_mov on geometry shaders based on cap

geometry shaders need to output this variable as well, and the variable
added using the pass on the vertex shader won't be passed through here

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5885>

radv: fix maxFragmentShadingRateRasterizationSamples

It's not a bitfield. This limit is purely informational.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8100>

radv: adjust the maximum number of coverage samples for VRS

It should actually be 4 because the maximum fragment size supported
by the hardware is 2x2.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8100>

swr: Fix crashes on Windows

SWR is missing implementation of pipe_context::flush_resource
function, which is now in the execution path on Windows.
This change adds an empty implementation (flush_resource
is NOOP in SWR) to prevent crashes

Reviewed-by: Krzysztof Raszkowski <krzysztof.raszkowski@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8159>

docs/features: update list of zink features

There's a few GL4 features that zink already supports.

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8158>

ac/llvm: insert phis before demote kill

LLVM (like NIR) requires phi instructions to be before any other
instructions in the block. ac_branch_exited() can insert non-phi
instructions before visit_block() adds phis, so visit_block() should add
phi instructions before the non-phi instructions ac_branch_exited()
inserts.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Fixes: aa757f4f8c3 ("ac/llvm: fix demote inside conditional branches")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8054>

doc/features: add VC4 driver

Include this Broadcom driver in the feature matrix, together with the
implemented extensions.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8137>

gitlab: add RADV bug report template

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7551>

radv: enable TC-compat HTILE for D32_SFLOAT+MSAA on GFX10+

This was disabled due to some depth/stencil resolve CTS failures
which are now fixed.

I figured that disabling TC-compat HTILE for D32_SFLOAT+MSAA reduced
performance in Control by -11% on Vega10. In fact, the game only uses
D32_SFLOAT for depth rendering.

This gives a huge boost in Control on Navi10 (eg. +17% in MSAA4x).
Note that the game is still slower than PRO without MSAA on Navi10,
but as fast (or even a bit faster) on Vega10.

I think TC-compat HILE could also be enabled for D32_SFLOAT_S8_UINT
but it needs more testing first.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8143>

turnip: fix cube map array image size calculation

imageSize() expects the last component of the return value to be the
number of layers in the texture array. In the case of cube map array,
it will return a ivec3, with the third component being the number of
layer-faces.

Fixes: dEQP-VK.image.image_size.cube_array.*

Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8087>

mesa: call Driver.TexParameter in glPopAttrib to fix r100, r200, old nouveau

The glPopAttrib optimizations incorrectly removed it.

Use GL_ALL_ATTRIB_BITS to mean "all texture parameters have changed" to
make it more efficient.

Fixes: d0e18550e26 - mesa: optimize saving/restoring bound textures for glPush/PopAttrib

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

mesa: don't restore texture state into unbound textures in glPopAttrib

It was a typo, or thinko, sort of.

Fixes: d0e18550e26 - mesa: optimize saving/restoring bound textures for glPush/PopAttrib

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

mesa: overallocate program parameter values

See the comment. This is something I spotted in the code. There is
no known bug caused by this.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

mesa: remove unused LastUniformIndex

This is now dead code.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

mesa: fix assertion paramList->LastUniformIndex < paramList->FirstStateVarIndex

This changes the code so that program parameters no longer have to be
sorted (meaning uniforms and constants are before state variables).

Instead of checking if the parameter is a state variable for every element,
teach all functions to handle non-state parameters safely. This is better
for the most common case where parameters are sorted or semi-sorted.

The new enum STATE_NOT_STATE_VAR identifes that a parameter is not
a state variable.

Fixes: 63f7d7dd - mesa: take advantage of sorted parameters in _mesa_load_state_parameters
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3914

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

draw: fix incorrect NIR support code

"state" contains NIR, while "vs->base.state" contains TGSI generated
from NIR. It was a typo.

This fixes the arb_vp subtest of: DRAW_USE_LLVM=1 piglit/bin/rasterpos

Fixes: df11ceaaaf7 - draw: add NIR support to draw_create_vertex_shader

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

nir_to_tgsi: fix NIR options instead of asserting

Drivers can return any NIR options, but nir_to_tgsi doesn't support
any options because it was written to assume certain options are set.

This fixes a regression caused when st/mesa switched to NIR for passing
shaders to the draw module.

Fixes: 44b7e1497f9 - st/mesa: don't generate TGSI for the draw VS because it now supports NIR too
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3949
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3965

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8046>

freedreno/a6xx: Flush depth at the end of bypass rendering, too.

We apparently don't have anything else making sure that it's flushed in
between use as a render target and use as a texture source, so bypass-mode
depth texture sampling could get stale data.

Fixes consistent (as far as I could see) failures in FD_MESA_DEBUG=nogmem
on:

dEQP-GLES31.functional.texture.multisample.samples_*.use_texture_depth_2d
dEQP-GLES31.functional.stencil_texturing.render.depth24_stencil8_draw

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8146>

mesa/st: Finalize the texture before BlitFramebuffer from it.

If the texture hasn't yet been validated, the current tex image contents
may have been ignored in favor of the texture's main miptree.  Fixes test
failure on softpipe and intermittent failures on radeonsi in:

https://gitlab.freedesktop.org/mesa/piglit/-/merge_requests/428

Cc: mesa-stable
Tested-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8091>

util/list: use helper function in list_is_singular

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

omx: use list_is_linked

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

r600: use list_is_linked

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

iris: use list_is_linked

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

gallium: use list_is_linked

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

nine: use list_is_linked

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

util/list: add list_is_linked

v2: verify node is valid in list_is_linked (Timothy)

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

iris: remove redundant check

list_del dereferences both next and prev, so if only one of them could
be NULL we would get crashes already.

Should fix "Dereference after null check" reported by Coverity.

Code was added in: 64b73b770b7 ("iris: Fix bad external BO hash table and zombie list interactions")

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

svga: remove duplicated code

list_del unconditionally sets "next" and "prev" to NULL.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8110>

nir/opt_if: split ALU from Phi more aggressively

If the only user is a trivial bcsel which in a second step
can be turned into a phi, this conversion is also worth it
even if the previous result is not undefined or constant.
Allows for some more loop unrolling or saves a few instructions.

Totals from 62 (0.04% of 139391) affected shaders (NAVI10):
SGPRs: 4976 -> 4992 (+0.32%)
VGPRs: 4408 -> 4472 (+1.45%); split: -0.45%, +1.91%
CodeSize: 453632 -> 464000 (+2.29%); split: -0.32%, +2.60%
MaxWaves: 527 -> 511 (-3.04%); split: +0.38%, -3.42%
Instrs: 84940 -> 86681 (+2.05%); split: -0.36%, +2.41%
Cycles: 11946844 -> 11783708 (-1.37%); split: -1.40%, +0.04%
VMEM: 9403 -> 10357 (+10.15%); split: +11.59%, -1.45%
SMEM: 3003 -> 3025 (+0.73%); split: +1.07%, -0.33%
VClause: 1756 -> 1997 (+13.72%); split: -0.11%, +13.84%
SClause: 2914 -> 2915 (+0.03%); split: -0.10%, +0.14%
Copies: 6426 -> 6768 (+5.32%); split: -4.14%, +9.46%
Branches: 2105 -> 2102 (-0.14%); split: -1.66%, +1.52%
PreSGPRs: 2921 -> 2909 (-0.41%); split: -0.55%, +0.14%
PreVGPRs: 4151 -> 4179 (+0.67%); split: -0.24%, +0.92%

cc: mesa-stable

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8123>

clover/api: Add extended versioning query for OpenCL C

The cl_khr_extended_versioning extension differs from the OpenCL 3.0
version on this specific as it only reports a single supported OpenCL C
version, whereas the OpenCL 3.0 equivalent will report all of them.

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7590>

clover/api: Add extended versioning query for built-in kernels

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7590>

zink: enable 64bit pipe caps

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: set nir options for 64bit handling based on feature presence

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: split ubo loading for 64bit types into 2x32bit loads

this loads 2x32 for each 64bit value and then zips it back into 64bit
for storage

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: more conversion ALUs

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: handle 64bit constant loading in ntv

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: add 64bit glsl basetype handling in ntv

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: change function params and asserts to permit 64bit types in ntv

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: set 64bit shader caps in ntv

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: add handlers for some bitfield ops in ntv

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: nir_op_b2f64 implementation

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: add some 64-bit conversion ALUs

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: support nir_op_f2f32

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

zink: add handling for 64bit values in spirv_builder

Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7654>

drisw: Prefer hardware-layered sw-winsys drivers over pure sw

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8128>

drisw: Add fallback logic for choosing a driver to use

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8128>

aco: add block to worklist in mark_block_wqm()

Since we're requiring the branch condition to be in WQM, we have to ensure
that the block is in the worklist.

Fixes Trials Fusion hang at 4K and High settings.

fossil-db (Sienna):
Totals from 216 (0.15% of 139391) affected shaders:
SGPRs: 13392 -> 13360 (-0.24%)
CodeSize: 1321184 -> 1318592 (-0.20%)
Instrs: 255310 -> 254662 (-0.25%)
Cycles: 2178360 -> 2174652 (-0.17%)

Affected fossils in fossil-db are dirt4, nier and youngblood.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3863
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8145>

zink: add struct type support for ntv

also interface types since they're the same

Erik Faye-Lund <erik.faye-lund@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8142>

zink: ntv formatting

Erik Faye-Lund <erik.faye-lund@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8142>

zink: handle struct derefs in ntv

Erik Faye-Lund <erik.faye-lund@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8142>

zink: add get_storage_class() ntv util

Erik Faye-Lund <erik.faye-lund@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8142>

zink: be more paranoid about array strides in ntv

arraystride is a required decoration for arrays of scalars, so ensure that
we put in some effort on this for the case where an array doesn't specify
an explicit stride

Erik Faye-Lund <erik.faye-lund@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8142>

ci/freedreno: Stop specifying the number of deqp threads

For "0" (its default) deqp-runner picks a number of jobs corresponding to
the CPU count, so set our hardware runners to use that (note that mesa's
deqp-runner.sh will pick a default of 4 if we don't specify a
DEQP_PARALLEL).

This means we'll allocate threads for the slow cores on a630 now, reducing
gles3 runtime from 6.5 minutes to around 5.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8132>

ci/freedreno: Treat all freedreno deqp runs as saving results.

They've been doing so since the webdav results upload was added.  This
means that we'll get normal truncated failures lists with the pointer to
the job artifacts, rather than filling a log file if you broke everything.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8132>

zink: add nir pass for splitting 64bit vertex attribs which cross slot boundaries

according to spec, dvec3 and dvec4 vertex attribs require 2 slots (locations),
and so the shader loads have to be explicitly split to reflect this

helpfully, gallium already gives us the vertex element state in a split format,
so no other changes are necessary to have this work as expected

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8141>

zink: support frem shader op

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8144>

tu: pCounterBuffers can be NULL in vkCmd*TransformFeedbackEXT()

According to the spec:

"pCounterBuffers is an optional array of buffer handles [...]
 If pCounterBuffers is NULL, then transform feedback will start
 capturing vertex data to byte offset zero in all bound transform
 feedback buffers."

"If counterBufferCount is not 0, and pCounterBuffers is not NULL,
 pCounterBuffers must be a valid pointer to an array [...]"

So counterBufferCount could be non-zero with pCounterBuffers
being NULL.

Fixes crash in RenderDoc when inspecting draw call with tesselation
or geometry shader present.

Fixes: 98b0d900 "turnip: rework streamout state and add missing counter buffer read/writes"
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8140>

tu: Ignore pTessellationState if there is no tesselation shaders

According to the spec:

"pTessellationState [...] is ignored if the pipeline does not
 include a tessellation control shader stage and tessellation
 evaluation shader stage."

Fixes crash in RenderDoc when inspecting draw call with
geometry shader but without tesselation shaders.

Fixes: eefdca2e "turnip: Parse tess state and support PATCH primtype"
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8140>

glsl/float64: make this compatible with glsl 330

just a minor tweak

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7483>

radv: add missing DB flush after depth/stencil resolve operations

I thought this was a bug in CTS but the Vulkan spec says:

    "VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT specifies write access
     to a color, resolve, or depth/stencil resolve attachment during
     a render pass or via certain subpass load and store operations."

So, VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT is used to synchronize
depth/stencil resolve attachments. Yes, it's counterintuitive.

This can't actually be fixed properly for now because RADV performs
the end subpass barrier *before* resolve attachments instead of after.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8138>

aco/ra: fix phi operand renaming

In case one operand was renamed and another operand came
from an incomplete phi, it could happen, that the original
name was not restored.

This has no impact on the code, but ensures correct SSA
is maintained during RA.

Cc: mesa-stable
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8109>

ci: Bump deqp to current vulkan-cts-1.2.5.0

This new version contains several fixes.

v1 (Eric):
 - Rebuild fail list for freedreno and lavapipe
 - Move freedreno crashes to skip list

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3902
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8088>

egl: fix EGL_EXT_protected_content/surface mixup

EGL_EXT_protected_surface introduces EGL_PROTECTED_CONTENT_EXT,
while EGL_EXT_protected_content is about protected context.

When I implemented EGL_EXT_protected_surface I mixed up the 2
names, so this commit fixes it.

Fixes: bd182777c8f ("egl: implement EGL_EXT_protected_surface support")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8122>

lavapipe: implement VK_EXT_index_type_uint8

Since Gallium supports 8 bit indices, this extension is a simple matter
of plumbing a value through, exposing a feature and flipping the switch
for the extension. This lets zink avoid up-converting the index-buffer
before drawing.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8082>

v3dv: fix BO list for TFU jobs

Instead of checking whether the source and destination are the same,
we should check if the underlying BOs are the same, since we may
be suballocating resources from the same allocation and the kernel
will fail to execute jobs if the BO list has duplicated entries.

Fixes aborts with Unreal Engine due to failed TFU jobs.

Fixes: 30f1fc25ce ('v3dv: implement TFU blits')
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8098>

v3dv: fix incorrect slice selection for TFU jobs

Fixes: 30f1fc25ce ('v3dv: implement TFU blits')
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8098>

symbol_table:fix mesa symbol table return scope error

According to the comment of this function,return a non-negative
number for the number of scopes between the current scope and
the scope where a symbol was defined.

Signed-off-by: cheyang <cheyang@bytedance.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8084>

glsl: Initialize glsl_type member name.

Fix defect reported by Coverity Scan.

Uninitialized pointer field (UNINIT_CTOR)
uninit_member: Non-static class member name is not initialized in
this constructor nor in any functions that it calls.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7766>

docs: Add a stub page for Panfrost

List the supported hardware with some remarks. This page should grow in
the future.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8096>

meson: add missing dependency on generated git_sha1.h

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8121>

docs: update calendar and link releases notes for 20.2.6

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8131>

docs: Add sha256 sums for 20.2.6

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8131>

docs: add release notes for 20.2.6

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8131>

pan/decode: Fix "Access to unknown memory" message formatting

The format string was missing a newline.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8130>

panfrost: Fix negative LOD bias support on Bifrost

The LOD bias can be negative, so mark it as signed in the XML.

The code in pan_cmdstream.c already calls FIXED_16 with
'allow_negative' set to true, so doesn't need to be adjusted.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8130>

pan/gen_pack: Fix signed integer packing

Any excess sign-extend bits would spill into adjacent fields, so mask
off anything after the end bit.

Shift from 2 instead of 1, because there needs to be one extra bit in
the mask as 'end' is inclusive.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8130>

panfrost: Split up batches with many jobs

If there are too many jobs in a batch, split it. Although the GPU
theoretically supports 65536 jobs in a batch, set the threshold lower
to avoid GPU timeouts.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8130>

zink: map some more formats

These formats correspond to missing vk_format_to_pipe_format formats, so
let's add support for them as well.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7982>