zink: fix device codegen extension detection

the logic for this was broken and failed to detect any extensions other
than the first one listed. instead, we must follow this logic chain:
1. check the extension name
  2a. if this is an extension that got promoted to core, check the @since version
    3a. if current version >= @since version
      4a. if the extension has required features/properties, check those
      4b. else set supported
    3b. else
      4a. if the extension has required features/properties, check those
      4b. else set supported
  2b. else
    4a. if the extension has required features/properties, check those
    4b. else set supported

Fixes: efe6f00e345 ("zink/codegen: do not enable extensions that are now core")

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9081>

nir/algebraic: Covert up-cast of down-cast to extract on Intel

This starts generating extract for bit sizes other than 32 but our
back-end handles that just fine.

Reviewed-by: Ian Romanick <ian.d.romanick@.intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>

nir/algebraic: Clean up up-cast of down-cast when we can

There are a bunch of cases where we can pretty quickly determine that
the high bits don't matter.  In these cases, delete the casts.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>

nir: Add some range analysis for used bits

This isn't 100% accurate, of course, but it should be good enough for
what we're about to do with it.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>

intel/nir: Lower 8-bit phis on Gen11+

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>

nir/lower_bit_size: Support phi instructions

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>

nir: Add a couple helpers for phis and cursors

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>

virgl: Return total video memory if available

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9009>

virgl: update headers

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9009>

frontends/va/config: Fix check for packed header config

Fixes: b4651890be4 ("frontends/va: Update conditional checks for code stability.")
Signed-off-by: Thong Thai <thong.thai@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4285
Reviewed-by: Leo Liu <leo.liu@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9020>

gitlab-ci: add intel APL and GLK devices with manual triggers

This add support for the Intel Apollo Lake and Gemini Lake families,
however the job will be disabled by default unless the developer
manually hit play for the iris-apl-traces and iris-glk-traces jobs in
GitLab CI.

These devices are still under experimental level support in
the Lava lab and are not guaranteed to work reliably yet. Once they
become reliable and more resilient we will enable them by default in
MesaCI.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8162>

gitlab-ci: build the iris gallium driver as well

For now we will start with iris. Our main focus is to enable the devices
to be used in MesaCI.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8162>

gitlab-ci: extend x86_64 kernel config to suport Intel devices

There are extra kernel config options we need to enable for Intel.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8162>

intel/compiler: Free resources on test teardown

Ensure that all resources are properly released by
properly parenting them to a memory context and releasing
the context during test teardown.

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8162>

intel/genxml: Free resource before exiting

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8162>

pan/bi: Use the correct size for UBO loads

Multiply by the destination bit size to get the number of bits to
load instead of assuming 32 bits.

Fixes: 2e57684d2d3 ("pan/bi: Implement load_ubo with the builder")
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9071>

radv: Do pipe misalignment check per plane.

Fixes: 4c99d6ff54b ("radv: flush L2 for images affected by the pipe misaligned issue on GFX10+")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9078>

broadcom/compiler: Merge instructions more efficiently

Instructions are allowed to access up to two rf registers, or one rf
register and a small immediate. This change allows qpu_merge_inst to
take full advantage of this by allowint the merging of two instructions
if they have no more than two different rf registers between them,
or one rf register and one small immediate. qpu_merge_inst rewrites
the instructions as needed to pack everything into raddr_a and raddr_b
in the merged instruction.

shader-db stats:
total instructions in shared programs: 19938769 -> 18929664 (-5.06%)
instructions in affected programs: 17929438 -> 16920333 (-5.63%)
helped: 95008
HURT: 242
helped stats (abs) min: 1 max: 785 x̄: 10.62 x̃: 7
helped stats (rel) min: 0.30% max: 21.25% x̄: 5.37% x̃: 4.98%
HURT stats (abs)   min: 1 max: 2 x̄: 1.10 x̃: 1
HURT stats (rel)   min: 0.30% max: 3.12% x̄: 1.62% x̃: 1.54%
95% mean confidence interval for instructions value: -10.67 -10.52
95% mean confidence interval for instructions %-change: -5.37% -5.33%
Instructions are helped.

total max-temps in shared programs: 3122664 -> 3112446 (-0.33%)
max-temps in affected programs: 124881 -> 114663 (-8.18%)
helped: 5445
HURT: 0
helped stats (abs) min: 1 max: 15 x̄: 1.88 x̃: 1
helped stats (rel) min: 1.49% max: 40.54% x̄: 8.97% x̃: 6.67%
95% mean confidence interval for max-temps value: -1.91 -1.84
95% mean confidence interval for max-temps %-change: -9.12% -8.81%
Max-temps are helped.

total sfu-stalls in shared programs: 38028 -> 41231 (8.42%)
sfu-stalls in affected programs: 6053 -> 9256 (52.92%)
helped: 664
HURT: 3380
helped stats (abs) min: 1 max: 2 x̄: 1.04 x̃: 1
helped stats (rel) min: 9.09% max: 100.00% x̄: 70.81% x̃: 100.00%
HURT stats (abs)   min: 1 max: 4 x̄: 1.15 x̃: 1
HURT stats (rel)   min: 0.00% max: 300.00% x̄: 46.39% x̃: 25.00%
95% mean confidence interval for sfu-stalls value: 0.76 0.82
95% mean confidence interval for sfu-stalls %-change: 25.03% 29.26%
Sfu-stalls are HURT.

total inst-and-stalls in shared programs: 19976797 -> 18970895 (-5.04%)
inst-and-stalls in affected programs: 17963129 -> 16957227 (-5.60%)
helped: 95017
HURT: 245
helped stats (abs) min: 1 max: 785 x̄: 10.59 x̃: 7
helped stats (rel) min: 0.30% max: 21.25% x̄: 5.35% x̃: 4.95%
HURT stats (abs)   min: 1 max: 2 x̄: 1.09 x̃: 1
HURT stats (rel)   min: 0.30% max: 3.12% x̄: 1.61% x̃: 1.54%
95% mean confidence interval for inst-and-stalls value: -10.64 -10.48
95% mean confidence interval for inst-and-stalls %-change: -5.35% -5.31%
Inst-and-stalls are helped.

v2 (Iago):
 - moved early return for naddrs > 2 even earlier.
 - only update {add,mul}.b mux if instruction has more than one operand.
 - don't OR b->raddr_{a,b} if we are not merging add/mul instructions.
 - don't initialize packed to 0.
 - minor style fixes.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9026>

meson: invalid keyword argument dependencies

Fixes: e4cc52c0 (vulkan: Add common extension tables)
Reviewed-by: Dylan Baker <dylan.c.baker@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8821>

radv: only apply the MRT output NaN fixup to non-meta shaders

We only want this workaround to be applied for game shaders.

Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4163
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/9048>

ci: Update baremetal kernel to 5.11 plus patches

For imx6 we need the following fec ethernet fix:
 c730ab423bfa ("net: fec: Fix temporary RMII clock reset on link up")

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9046>

lavapipe: add support for missing 10/10/10/2 formats.

The snorm formats don't work for blitting, so don't expose
them for that.

This passes all CTS tests that it enables.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8907>

lavapipe: add support for 2/10/10/10 scaled formats.

These will be used for zink CI testing.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8907>

llvmpipe: don't support scaled formats outside vertex buffers

Scaled formats are usually only needed as vertex buffer formats,
don't expose them as supported for other things.

Lavapipe will use this to export the correct formats support.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8907>

util/format: add helper to check if a format is scaled.

lavapipe/llvmpipe need this to rule out scaled formats for non-vertex
format usage. Note NONE is defined as scaled in u_format, but for
the purposes of this helper I don't want it to be considered scaled.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8907>

zink: support nir_intrinsic_group_memory_barrier

needed for cts

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9073>

features: mark off GL 4.6 and ES 3.1 for zink

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: GLSL 460

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: PIPE_CAP_GL_SPIRV

still needs a bunch of fixing, but 75% of tests pass

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: enable pipeline statistics cap

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: enable PIPE_CAP_QUERY_SO_OVERFLOW

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: enable PIPE_CAP_POLYGON_OFFSET_CLAMP

this is already handled

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: enable PIPE_CAP_DRAW_PARAMETERS

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: enable PIPE_CAP_TGSI_VOTE

ARB_shader_group_vote

Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9072>

zink: add util function for submitting the compute batch

sometimes we just want to submit this and not wait on it

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9063>

zink: rewrite drawid based on shader key value

same as basevertex

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: break out push constant creation in compiler and add drawid value

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: add a vs shader key for rewriting gl_DrawID

if we get a direct draw, we have to rewrite the drawid for it to work as expected

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: add a draw_id param to vs push constants

gallium always gives us multidraws as individual draws, so we need to feed
this through to the shader

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: wrap shader gl_BaseVertex access with a bcsel based on push constant state

vulkan has different mechanics than gl for this variable based on whether the
current draw is indexed, so we need to rewrite the access here for that case

this also requires that we add some padding to the tcs shader injection to
account for new members being added to the push constant

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: add push constant value to indicate whether the current draw is indexed

due to semantic differences between gl and vk variables, this is going to be
necessary for gl_BaseVertex

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: rework tcs injection to be more compatible with new push const struct

we can simplify the push constant loader to directly take the struct member
index here and then pass that directly to simplify things for future use

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: create a struct for tracking push constant layout

it'll be helpful to update this as we add more values here so we can
avoid having to update all the users

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: add handling for ARB_shader_draw_parameters variables in ntv

* gl_DrawID
* gl_BaseVertex
* gl_BaseInstance (was already semi-handled)

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8971>

zink: handle 1bit undef values in ntv

spirv requires that 1bit values be bool types, not uints

Fixes: 93af00502eb ("zink: use uvec for undefs")

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9059>

zink: fix slot mapping for legacy gl io with tess stages

e.g., gl_BackColor is a thing

Fixes: aec03553e1b ("zink: fix tess shader i/o variables")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9019>

zink: add support for pipeline statistics queries

so easy!

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9010>

zink: hook up cs invocation queries to the compute batch

compute batches are separate, so we have to run these queries on the
appropriate batch to see the expected results

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9010>

zink: unset generated TCS if its parent TESS is unset

ensure this doesn't get pulled in during the next program update

Fixes: 334759d8509 ("zink: implement passthrough tcs shader injection")

Reviewed-by: Hoe Hao Cheng <haochengho12907@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9057>

aco: fix assertion in insert_exec_mask pass

Fixes: a56ddca4e80a6ef7bb0c44edb4e5b6169510aaca ('aco: make all exec accesses non-temporaries ')
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9047>

aco: fix transition_to_{WQM,Exact} if exec.back() is not in exec

This can happen at merge blocks.

fossil-db (GFX10.3):
Totals from 25229 (17.25% of 146267) affected shaders:
CodeSize: 58575920 -> 58571376 (-0.01%); split: -0.01%, +0.00%
Instrs: 10979245 -> 10978109 (-0.01%); split: -0.01%, +0.00%
SClause: 591817 -> 591816 (-0.00%)
Copies: 604987 -> 603851 (-0.19%); split: -0.19%, +0.00%
Cycles: 96088796 -> 96084252 (-0.00%); split: -0.00%, +0.00%
VMEM: 10470372 -> 10470368 (-0.00%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: a56ddca4e80 ("aco: make all exec accesses non-temporaries")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4299
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9047>

mesa: add debug code to catch missing _mesa_update_valid_to_render_state calls

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: inline draw validate functions

This just moves the functions to draw.c.

Our CPU overhead is so low now that this increases performance by a few %.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: inline _mesa_set_draw_vao and set_varying_vp_inputs for draw calls

This just moves the functions to draw.c. It's inlined because the function
name disappears in the profiler. It improves performance.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: gather errors and call _mesa_error only once in validate_Draw

This mainly skips all the jumping around various _mesa_error calls that
are never taken. There is a tiny measurable improvement in CPU overhead.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: precompute draw time determination of enabled vertex arrays

Another draw time optimization that precomputes the value only when
_VPMode is changed.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: precompute _mesa_get_vao_vp_inputs

It's called for every draw, so this is important.

All inputs of _mesa_get_vao_vp_inputs are changed when
update_attribute_map_mode is called, so we can just compute
the value there.

The assertion ensures correct behavior in debug builds.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: set _DrawVAOEnabledAttribs only when it changes

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move gl_context::varying_vp_inputs into ctx->VertexProgram._VaryingInputs

This is where it belongs.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: optimize set_varying_vp_inputs by precomputing the conditions

set_varying_vp_inputs is called every draw call, which checks
_Maintain*Program. Let's move that checking out of there.

This adds a new flag that determines whether set_varying_vp_inputs
should do anything.

All code that changes _Maintain*Program must now reinitialize the new
flag. This is done by new function _mesa_reset_vertex_processing_mode.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: validate numInstances in common functions to unify code

Just a cleanup.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move disallowed TFB in DrawElements on GLES from draws to state changes

glDrawElements is disallowed by GLES when transform feedback is active
and GS is not supported.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: add a separate valid primitive mask just for glDrawElements

No change in behavior.

This will be used for draw validation required by glDrawElements but not
glDrawArrays.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: don't skip draws with count == 0 or numInstances == 0

These are checked by drivers (all gallium drivers and
_mesa_draw_gallium_fallback), so they are redundant here.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: skip MultiDrawArrays with primcount == 0

This is the only zero parameter that gallium can't handle.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: remove an optional GL error about mapped buffers during execution

Not having this here, even if the branch is not taken, increases
CPU performance by 2% on radeonsi. If some drivers need this, the spec
does allow GL termination, meaning abort(), which is a more effective
alternative given that this never happens.

You may ask, do we really pay a 2% performance hit for every conditional
not taken? For some of them, we do.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: call _mesa_update_state() before validation

It's called in both the no_error path and the draw validation path, so we
might as well call it before the no_error conditional.

This decreases CPU overhead a lot, which I'm unable to explain. It seems
to be random gcc behavior, which I've seen quite a lot now.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: remove optional draw validation code to increase performance

Some drivers do similar validation in their draw calls.

If drivers need some of this, we can do it in the drivers or even in
_mesa_draw_gallium_fallback for classic drivers.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: remove VERBOSE_DRAW

Let's declutter the code.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: optimize the dual source blend error checking using a bitmask

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: inline _mesa_valid_to_render now that it doesn't do validation

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move blending validation from draws to state changes

This is a step towards removing _mesa_valid_to_render.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move GL_FILL_RECTANGLE validation from draws to state changes

This is a step towards removing _mesa_valid_to_render.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move ARB program and integer FBO validation from draws to state changes

This is a step towards removing _mesa_valid_to_render.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move FBO completeness checking from draws to state changes

_mesa_update_framebuffer_visual already calls _mesa_update_valid_to_render-
_state, so we just need to call it where FBOs are marked incomplete.

This is a step towards removing _mesa_valid_to_render.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move some uniform debug code from draws to state changes

This is a step towards removing _mesa_valid_to_render.

I don't know what this code does. This commit probably breaks it, but it
doesn't seem important. We could remove the code.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move sampler uniform validation from draws to state changes

This is a step towards removing _mesa_valid_to_render.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move shader pipeline validation from draws to state changes

This is a step towards removing _mesa_valid_to_render.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: don't report 1 for GL_VALIDATE_STATUS if user didn't validate pipeline

dEQP expects GL_VALIDATE_STATUS to be set to 0 if the user didn't call
glValidateProgramPipeline even if the pipeline is valid, but we will
validate pipeline objects in state changes, not draw calls.

This will fix a failure in this test caused by the next commit:
dEQP-GLES31.functional.state_query.program_pipeline.validate_status_get_program_pipelineiv

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: add skeleton code for DrawPixels/CopyPixels/Bitmap precomputed validation

This is for precomputing draw time validation in state changes.
The next commit will use this.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: inline check_valid_to_render

because it has only one use.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: fold most of check_valid_to_render into _mesa_update_valid_to_render_state

Set the mask (ValidPrimMask) to 0 if draw calls should generate
GL_INVALID_OPERATION. The mask is updated when states are changed.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: move check_valid_to_render call into _mesa_valid_prim_mode

It's going to be mostly eliminated by ValidPrimMask, so let's put them
together.

This also enables check_valid_to_render for glBegin, which seems to have
been omitted accidentally.

_mesa_valid_to_render is now always called by _mesa_valid_prim_mode.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: precompute draw time prim validation during state changes

This moves the body of _mesa_valid_prim_mode into new function
_mesa_update_valid_to_render_state, which is called when the affected
states are changed and sets new variable gl_context::ValidPrimMask,
which determines errors reported by draw calls.

_mesa_valid_prim_mode only has to check ValidPrimMask and choose
between GL_INVALID_ENUM and GL_INVALID_OPERATION depending on whether
the primitive would be allowed by the GL version and extensions
(GL_INVALID_OPERATION) or not (GL_INVALID_ENUM).

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: precompute all valid primitive types at context creation

New variable gl_context::MaxValidPrimMask is set at context creation
and determines the valid primitive types for that context.

The Max prefix indicates that the mask doesn't mask out primitives
disallowed by current states and shaders.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

mesa: optimize draw index type checking

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8798>

freedreno: Add missing dep on freedreno tracepoints.

We were only get guaranteed that libfreedreno (and thus the tracepoints
generation) was ready when we linked, not when we compiled the gmemtool.c
that also used it.

Fixes: a02dcb970fff ("freedreno: Add GPU tracepoints")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9056>

vulkan: document flags choice for vkGetDeviceQueue

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9053>

ci/v3d: Add V3D and V3DV testing

Add OpenGL and Vulkan testing for V3D and V3DV respectively.

Add also a couple of manual piglit jobs for V3D.

v2:
 - Replace custom mustpass with running fraction of tests (Eric)

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8745>

ci: add option to overwrite CPU arch

When loading Vulkan ICD file, it uses the CPU machine identifier to
load the correct one, in case multiple versions are installed.

This is fine if the machine where Mesa has been built and the machine
where the test is run are exactly the same. But this is not always the
case. As example, for armhf architecture, the machine where Mesa is
built is identified as `arm7hlf`, but the Raspberry Pi 4 is identified
as `armv7l`, so it will fail to load the ICD file, though both are
totally compatible.

This allow to define the architecture instead.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8745>

aco: add DeviceInfo

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

aco: consider that GFX10.3 allocates LDS in 1024 byte blocks

fossil-db (GFX10.3):
Totals from 3 (0.00% of 139391) affected shaders:
VMEM: 513 -> 511 (-0.39%)
SMEM: 94 -> 92 (-2.13%)
VClause: 31 -> 30 (-3.23%)

fossil-db (GFX10.3, wave32):
Totals from 4 (0.00% of 139391) affected shaders:
VClause: 82 -> 81 (-1.22%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

radv,aco: add radv_nir_compiler_options::wgp_mode

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

aco: add Program::wgp_mode

Instead of assuming WGP mode on GFX10+ in different places, add a member
to Program that can be used instead.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

aco: fix waves calculation for wave32

fossil-db (GFX10.3, wave32):
Totals from 176 (0.13% of 139391) affected shaders:
SGPRs: 16648 -> 16640 (-0.05%)
VGPRs: 18920 -> 19076 (+0.82%); split: -0.30%, +1.12%
CodeSize: 2354172 -> 2354288 (+0.00%); split: -0.01%, +0.01%
MaxWaves: 1618 -> 1627 (+0.56%); split: +0.68%, -0.12%
Instrs: 435756 -> 435761 (+0.00%); split: -0.02%, +0.02%
Cycles: 8858360 -> 8869960 (+0.13%); split: -0.01%, +0.14%
VMEM: 55899 -> 57220 (+2.36%); split: +2.53%, -0.17%
SMEM: 10323 -> 10374 (+0.49%); split: +0.73%, -0.23%
VClause: 8307 -> 8290 (-0.20%); split: -0.24%, +0.04%
SClause: 16573 -> 16577 (+0.02%); split: -0.01%, +0.03%
Copies: 24641 -> 24652 (+0.04%); split: -0.24%, +0.28%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

radv: round up max_lds_per_simd / lds_per_wave

If each SIMD has to get an different number of waves, report the maximum.

One example of a situation is when a single-wave workgroup uses more than
max_lds_per_simd. This change causes radv_get_max_waves() to report a
single wave per SIMD instead of none.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

radv: use lds_{encode,alloc}_granularity

This fixes a issue in radv_get_max_waves() where it aligned the LDS
allocation to 512 bytes instead of 1024 on GFX10.3.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

ac: split lds_granularity into encode and allocation granularities

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

radv: switch MaxWaves statistic to wave32 waves

Always return the wave32 waves instead of wave64 waves because the wave32
wave count is more precise in the case of wave32.

This also fixes usage of lds_per_wave in wave32.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>

radv: fix max_lds_per_simd on GFX10

num_simd_per_compute_unit was the number of SIMDs per compute unit, but
lds_size_per_workgroup was the bytes of LDS per WGP.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8761>