spirv: reverse order in matrix multiplication

This will create code that is easier to combine into MADs/FMA when the
last component of the vector is 1.0.

nir_opt_algebraic_late has an optimization to do something similar but it
only works for inexact code, if the multiplication-by-1 optimization is
done before it and if the backend enables fuse_ffma.

fossil-db (Navi):
Totals from 4296 (3.75% of 114665) affected shaders:
SGPRs: 283468 -> 283764 (+0.10%); split: -0.02%, +0.12%
VGPRs: 172868 -> 172904 (+0.02%); split: -0.09%, +0.11%
CodeSize: 14045312 -> 14027128 (-0.13%); split: -0.15%, +0.02%
MaxWaves: 59285 -> 59282 (-0.01%); split: +0.04%, -0.05%
Instrs: 2703507 -> 2683187 (-0.75%); split: -0.76%, +0.00%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5631>

nir: scalarize fdot in reverse

This will create code that is easier to combine into MADs/FMA when the
last component is 1.0.

nir_opt_algebraic_late has an optimization to do something similar but it
only works for inexact code, if the multiplication-by-1 optimization is
done before it and if the backend enables fuse_ffma.

fossil-db (Navi):
Totals from 85583 (74.64% of 114665) affected shaders:
SGPRs: 4556060 -> 4558596 (+0.06%); split: -0.07%, +0.12%
VGPRs: 3315060 -> 3312984 (-0.06%); split: -0.23%, +0.17%
SpillSGPRs: 13552 -> 13553 (+0.01%)
CodeSize: 184962756 -> 184431388 (-0.29%); split: -0.32%, +0.03%
MaxWaves: 1208693 -> 1209361 (+0.06%); split: +0.17%, -0.11%
Instrs: 35678819 -> 35361617 (-0.89%); split: -0.91%, +0.02%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5631>

panfrost: Set .array_size on Bifrost

Fixes dEQP-GLES2.functional.texture.mipmap.cube.generate.rgba8888_nicest

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Don't emit TEXS for array textures

No place for the extra coordinate.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Handle 3D/array coordinates

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Track tex data register swizzles

So we can pass through a .z component.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Add bi_emit_array_index helper

Need to handle type conversion.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Drop unused swizzles

Missed during format cleanup.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Advertise Bifrost support

Drop the PAN_MESA_DEBUG=bifrost flag. Load on Bifrost chips by default.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Disable point sprites on Bifrost

It doesn't look like the lowering in !6473 will land before the branch
point. Let's nop out point sprites in the backend to avoid MMU faults
from creating invalid Midgard-style varyings.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Lower +CUBEFACE2

We need to inject a *CUBEFACE1 at pack-time so everything works out.
This is a pretty ugly hack but it'll hold us over until we have a real
scheduler, at which point it won't be necessary at all.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Suppress disassembly for internal shaders

Backport of 756441b2979c2664aaa13a67dbdaf5b7e853286e for bifrost.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Lower cube map coordinates

We need to do the transform specified in the OpenGL spec ourselves, with
some assistance from the hardware.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Hook up cube instructions packing

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Split special class in two

Some special instructions are scheduled on the FMA unit, let's add a
new class for this case and rename the old one accordingly.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: Move special instruction packing to a separate helper

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

pan/bi: s/t0/t1/ in bi_disasm_dest_add()

The ADD unit stores its result in t1 not t0.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Implement v7 texture payloads

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Add array size to XML

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Suppress Bifrost prefetching

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

panfrost: Leave push_constants pointer to NULL if there's no uniform

This removes a warning in pandecode.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7408>

mesa: do not throw _mesa_problem when invalid enum is used

Like with other getters, invalid enum is dealt in find_value by setting
error to GL_INVALID_ENUM and returning INVALID_TYPE which makes
get_value_size return 0.

Fixes false 'implementation errors' seen with Piglit test:
   ext_external_objects-memory-object-api-errors

   "Mesa 20.3.0-devel implementation error: invalid value type in GetUnsignedBytei_vEXT()
   Please report at https://gitlab.freedesktop.org/mesa/mesa/-/issues"

v2: add assert to get_value_size() (Lionel)

Fixes: e064d660205 ("mesa: implement glGetUnsignedByte{v|i_v}")
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Eleni Maria Stea <estea@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7403>

mesa/st: call memobj_destroy only if there is memory imported

Something may go wrong during import which leaves pointer to null and
when ctx and it's shared state gets destroyed we will attempt to call
memobj_destroy. Instead of forcing every driver to handle it, add check
here.

Fixes crashes with Piglit test:
   ext_external_objects_fd-memory-object-api-errors

Fixes: 99cf9108340 ("mesa/st: Actually free the driver part of memory objects on destruction.")
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Eleni Maria Stea <estea@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7403>

aco: select v_mul_{hi}_u32_u24 for 24-bit multiplications

This is based on the NIR range analysis. v_mul_u32_u24 is VOP2, while
v_mul_lo_u32 is VOP3, so that should reduce codesize.

fossils-db (Vega10):
Totals from 12590 (9.22% of 136546) affected shaders:
SGPRs: 680207 -> 677271 (-0.43%); split: -0.47%, +0.04%
VGPRs: 620840 -> 620856 (+0.00%); split: -0.02%, +0.02%
CodeSize: 37930200 -> 37774088 (-0.41%); split: -0.41%, +0.00%
Instrs: 7463550 -> 7458120 (-0.07%); split: -0.07%, +0.00%
Cycles: 133487628 -> 133427532 (-0.05%); split: -0.05%, +0.00%
VMEM: 2514729 -> 2513426 (-0.05%); split: +0.02%, -0.08%
SMEM: 1533579 -> 1532795 (-0.05%); split: +0.05%, -0.10%
VClause: 231391 -> 231389 (-0.00%); split: -0.01%, +0.00%
SClause: 255352 -> 255294 (-0.02%); split: -0.04%, +0.02%
Copies: 605821 -> 600352 (-0.90%); split: -0.92%, +0.02%
Branches: 133739 -> 133743 (+0.00%); split: -0.00%, +0.00%
PreSGPRs: 351092 -> 348048 (-0.87%)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7405>

aco: store NIR range analysis data to the isel context

It will be used to optimize some ALU instructions.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7405>

zink: clamped maxPerStageDescriptorUniformBuffers limits to INT_MAX when stored as uint32_t.

This stops issues if a driver returns a value that is greater than a signed int.
Also make it match many of the other limit versions conversions.
Seen on Radeon drivers, IIRC. gpuinfo.org also reports many GPUs returning 4GB values.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Erik Faye-Lund <erik.faye-lund@collabora.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7402>

radv: fix shader caching with NaN fixup workaround

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: 6f21995f986 ("radv: add new drirc option radv_enable_mrt_output_nan_fixup")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7423>

radv: fix shader caching with discard->demote workaround

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: bdd75874144 ("radv: use nir_lower_discard_to_demote to work around game bugs")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7423>

radv: add some missing radv_{start,stop}_feedback

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/7339>

android: freedreno: Add freedreno_dev_info.[ch] to Makefile.sources

Addresses the following linker error when building for Android:

    ld.lld: error: undefined symbol: freedreno_dev_info_init
    >>> referenced by freedreno_screen.c:1001 (external/mesa3d/src/gallium/drivers/freedreno/freedreno_screen.c:1001)
    >>>               freedreno_screen.o:(fd_screen_create) in archive [..]/libmesa_pipe_freedreno_intermediates/libmesa_pipe_freedreno.a

These functions were introduced in a file that was not included in the
Android build yet.  Also sort the list of files alphabetically as
requested in an earlier MR.

Fixes: 4a0bdf47e43 ("freedreno: Introduce common device info struct")
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7411>

broadcom/compiler: remove v3d_fs_key depth_enabled field.

It is not used right now, so keeping it adds some noise/confusion.

So far configuring Z test are done through the CFG_BITS. See
v3dX(emit_state) at v3dx_emit.c for v3d, and pack_cfg_bits at
v3dv_pipeline.c for v3dv. There flags like z_updates_enable and others
are filled up.

That key field seems like a leftover coming from using vc4 as
reference, as that driver defines and uses a field with name name.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7421>

intel/compiler: remove branch weight heuristic

As a result of this patch, compiler chooses SIMD32 shaders more
frequently.

Current logic is designed to avoid regressions from enabling SIMD32 at
all cost, even though the cases where regression can happen are probably
for smaller draw calls (far away from the camera and though smaller).

In Intel perf CI this patch improves FPS in:
- gfxbench5 alu2:      21.92% (gen9), 23.7%  (gen11)
- synmark OglShMapVsm:  3.26% (gen9),  4.52% (gen11)
- gfxbench5 car chase:  1.34% (gen9),  1.32% (gen11)
No observed regressions there.

In my testing, it also improves FPS in:
- The Talos Principle:   2.9% (gen9)

The other 16 games I tested had very minor changes in performance
(2/3 positive, but not significant enough to list here).

Note: this patch harms synmark OglDrvState (which is not in Intel perf
CI) by ~2.9%, but this benchmark renders multiple scenes from other
workloads (including OglShMapVsm, which is helped in standalone mode)
in tiny rectangles. Rendering so small drastically changes branching
statistics, which favors smaller SIMD modes. I assume this matters
only in micro-benchmarks, as in real workloads more expensive (with
more uniform branching behavior) draw calls dominate.

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Acked-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7137>

intel/compiler: use C++ template instead of preprocessor

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7382>

Revert "gallium/dri: fix dri2_from_planar for multiplanar images"

It breaks some assumptions in the iris driver, leading to crashes.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Tested-by: Jordan Justen <jordan.l.justen@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7391>

tu: Support rasterizerDiscardEnable and RasterizationStreamSelect

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6962>

tu: Support geometryStreams

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6962>

util/bitset: Add a range iterator helper

I need this for emitting the SO program for turnip, where we want to
skip over unused slots by manually advancing the counter. freedreno will
also want to use it when it supports multistream streamout.

Reviewed-by: Rob Clark <robdclark@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6962>

ir3: Support geometry streams

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6962>

freedreno/a6xx: Update SO registers for streams

These seem to be unchanged from a5xx, so a5xx could probably be updated
too.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6962>

zink: add VK_STRUCTURE_TYPE_WSI_MEMORY_ALLOCATE_INFO_MESA for WSI allocations

Since Zink doesn't use swapchains to create presentable images, drivers
lose the capacity to identify memory allocations for them, which is a problem
when the underlying platform has special requirements for these, such as
needing to allocate them on a particular device. Including this struct in the
pNext chain, which is the same thing that the Mesa Vulkan WSI code does when
allocating memory for swapchain images, gives drivers a chance to identify
and handle these memory allocations properly.

v2: follow Zink's conventions for pNext chains (Mike)
v3: add scanout parameter for VkImage creation (Daniel)
v4: don't add a dependency on vulkan util (Erik)
v5: include vulkan directory for Zink builds

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com> (v2)
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7378>

etnaviv: move etna_destroy_shader(..) to generic location

Before this change we had two identical etna_destroy_shader
functions - one for tgsi and one for nir.

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/7397>

etnaviv: move etna_dump_shader(..) to generic location

Before this change we had two identical etna_dump_shader
functions - one for tgsi and one for nir.

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/7397>

etnaviv: convert from tgsi semantic/index to varying-slot

Prep work to unify some tgsi and nir compiler functions.
No deqp and piglit regressions.

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/7397>

nir: make tgsi_varying_semantic_to_slot(..) public

I want to use this function for etnaviv's TGSI compiler.

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/7397>

amd/llvm,aco: Replace VLA with alloca

MSVC will never support VLA, so use alloca instead.

Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7157>

zink: require Vulkan timestamp queries for time query caps

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7374>

tgsi: Initialize tgsi_declaration_dimension padding.

Silence this Coverity defect.

Uninitialized scalar variable (UNINIT)
uninit_use: Using uninitialized value dim. Field dim.Padding is uninitialized.

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/7300>

gallium: Remove duplicate resource variable.

Fix defect reported by Coverity Scan.

Evaluation order violation (EVALUATION_ORDER)
write_write_typo: In resource = resource =
ntt_ureg_src_indirect(c, ureg_src_register(TGSI_FILE_IMAGE, 0U),
instr->src[0]), resource is written twice with the same value.

Fixes: 34cc6a804ec9 ("gallium: Add a nir-to-TGSI pass.")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7326>

nir/validate: Explain why we don't use nir_foreach_block

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7409>

zink: always reset query pools on next query begin

this ensures we pull in any flushes that are about to happen if we're
ending the query during a batch flush, which helps us roll over our pending
results

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: always use query->type for starting/stopping xfb queries

we're going to be seeing some overlap here

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: rework query overflow handling

this adds a query field to denote the last point at which a query was api started,
which is then used every time we call in to get_query_result as the starting point

this is important when we want to be able to return the same result set multiple
times

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: fixup gs/xfb tracking for primitives generated queries

need to index by query id here to ensure correct usage

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: store batch id onto query object at time of start

this is useful for knowing immediately whether a query has results available

for time queries, this ends up being end_query() since that's when the timestamp
is written

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: also create an xfb query for every primitives generated query

this query behaves differently when xfb is activated, specifically with
regard to vertex streams. it's super clunky, but we need to actually run
both queries and use results based on whether xfb was active during the
query

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: more correctly handle PIPE_QUERY_PRIMITIVES_GENERATED queries

in normal operation we want to be using INPUT_ASSEMBLY_PRIMITIVES_BIT,
but then when we break out the geometry shaders we actually want to
be using GEOMETRY_SHADER_PRIMITIVES_BIT, which means we need to track
whether a query has a gs active for draws

to do this, we keep a list of all these queries with this type and
iterate over it every draw to flag the gs state of the query that's
being drawn to. this works because our ring buffer of batches will
always wait on a fence after a full cycle, meaning there can only ever
be 4 queries with outstanding results

Fixes: e40a77ea5d0 ("zink: use right vulkan type for GL_PRIMITIVES_GENERATED queries")

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

zink: deduplicate some query result code

no functional changes, just a small simplification

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7195>

lavapipe: use resource get param.

This uses the resource get param to get proper values for image
subresource layouts.

Fixes:
dEQP-VK.image.subresource_layout*

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6639>

llvmpipe: add resource get param support.

This implements this resource get param callback and uses it
to return image information for lavapipe.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6639>

gallium: add a layer stride pipe resource parameter.

This will be used by llvmpipe and lavapipe

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6639>

gallium: add a level parameter to resource parameter get

For lavapipe interface to llvmpipe there is a need to retrieve per-level
info, so this seems like the best interface to use for it.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6639>

zink: call the reset callback not only during a status check

Fixes: 95b9fc41 ("zink: implement pipe_device_reset_callback")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7404>

i965: remove prototypes of not-existing functions

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7353>

intel: remove dead code

Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7353>

anv: always annotate memory returned from anv_gem_mmap

anv_bo_pool_alloc expects that the memory returned by and_gem_mmap
was annotated using VALGRIND_MALLOCLIKE_BLOCK, but anv_gem_mmap_offset
didn't do that. Move annotation from anv_gem_mmap_legacy to common
code.

Fixes: 4abf0837cdb ("anv: Add support for new MMAP_OFFSET ioctl.")

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/7381>

turnip: rework android gralloc path so it doesn't call tu_image_create

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7406>

os: Fix open result check.

Fix defect reported by Coverity Scan.

Argument cannot be negative (NEGATIVE_RETURNS)
negative_returns: f is passed to a parameter that cannot be negative.

CID: 1364709
Fixes: 13fa0513569d ("auxiliary/os: add new os_get_command_line() function")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7344>

iris/bufmgr: Handle NULL bufmgr in iris_bufmgr_get_for_fd

iris_bufmgr_create can return NULL, in which case we'd crash in
list_addtail.

Reported by Coverity/clang (for i965, but iris code looks the same).

Fixes: 7557f1605968 ("iris: share buffer managers accross screens")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7335>

i965/bufmgr: Handle NULL bufmgr in brw_bufmgr_get_for_fd

brw_bufmgr_create can return NULL, in which case we'd crash in
list_addtail.

Reported by Coverity/clang.

Fixes: 4094558e8643 ("i965: share buffer managers across screens")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7335>

freedreno: Use freedreno_dev_info

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7385>

tu: Use freedreno_dev_info

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7385>

freedreno: Introduce common device info struct

This will collect all the various alignments, sizes, and magic values
and set them appropriately, replacing the various pieces scattered
throughout the drivers.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7385>

nir/large_constants: only search for constant duplicates

Fixes: b6d47535684 ("nir/large_constants: De-duplicate constants")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3706
Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7350>

nir: Stabilize compact_components sort

Incorporate location_frac into qsort comparison. qsort is not required
to be stable, and MSVC implementation is not.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7399>

docs/features: add Vulkan 1.2

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6258>

docs/features: update unpromoted Vulkan extensions

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6258>

nir/lower_returns: Append missing phis' sources after "break" insertion

After we lowered `return` into `break` - the control flow is changed and
the block with this change has a new successor, which means that in this
new successor phis should have additional source.

Since the instructions that use phis in the successor are predicated -
it's ok for a new phi source to be undef.

If `return` is lowered in a nested loop, `break` is inserted in the outer
loops, so all new blocks with break require the same changes to phis
described above.

Examples of NIR before lowering:

  block block_0:
  loop {
     block block_1:
     if ssa_2 {
       block block_2:
       return
       // succs: block_6
     } else {
       block block_2:
       break;
       // succs: block_5
     }
     block block_4:
  }
  block block_5:
  // preds: block_3
  vec1 32 ssa_4 = phi block_3: ssa_1
  // succs: block_6
  block block_6:

Here converting return to break should add block_2 to the phis
of block_5.

 block block_0:
 loop {
    block block_1:
    loop {
       block block_2:
       if ssa_2 {
         block block_3:
         return
         // succs: block_8
       } else {
         block block_4:
         break;
         // succs: block_6
       }
       block block_5:
    }
    block block_6:
    break;
    // succs: block_7
 }
 block block_7:
 // preds: block_6
 vec1 32 ssa_4 = phi block_6: ssa_1
 // succs: block_8
 block block_8:

Here converting return to break will insert conditional break in
the outer loop, changing block_6 predcessors.

Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3322
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3498
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6186>

zink: implement pipe_device_reset_callback

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6763>

v3dv: fix occlusion query inheritance in secondary command buffers

If a secondary command buffer has occlusion query inheritance then
draw calls recorded in it should update an active occlusion query
counter started in the primary command buffer.

If executing the secondary in a primary required to emit jobs and
not just a branch instruction, then we might need to create a new
job for the primary as well, and in that case we would lose the
occlusion query state, so we need to re-emit it at that point so
any additional draw calls recorded into the secondary that is being
executed continue to update the counter.

Fixes:
dEQP-VK.query_pool.concurrent_queries.secondary_command_buffer

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7373>

v3dv: add support for timestamp queries

V3D doesn't provide any means to acquire timestamps from the GPU
so we have to implement these in the CPU.

v2: enable timestampComputeAndGraphics and set timestampPeriod (Piñeiro)

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7373>

radeonsi: fix RADEON_FLUSH flags conflicts

RADEON_FLUSH_TOGGLE_SECURE_SUBMISSION and RADEON_FLUSH_NOOP used the same value.

Fixes: ed3c5fe4694 ("radeonsi: implement GL_INTEL_blackhole_render")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

driconf: add disable_protected_content_check option

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

egl/dri2: implement createImageFromDmaBufs3

And refuse to import image with protected_content enabled.

We don't want a compositor to import an encrypted buffer in a image
without the ProtectedContent attribute enabled, because that will
lead to incorrect display.

Similarly, if the compositor thinks the image is encrypted, we fail
the import if the buffer is not.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

dri: introduce createImageFromDmaBufs3

Extends createImageFromDmaBufs2 with a protected_content flag.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

egl: handle EGL_PROTECTED_CONTENT_EXT for eglImage

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

radeonsi: enable PIPE_CAP_DEVICE_PROTECTED_CONTENT

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

egl: implement EGL_EXT_protected_surface support

Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

radeonsi: honor PIPE_BIND_PROTECTED

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

gallium: introduce PIPE_BIND_PROTECTED

Resources using this flag will be encrypted (eg using TMZ on radeonsi).

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

gallium: add new cap PIPE_CAP_DEVICE_PROTECTED_CONTENT

Will be used to implement EGL_EXT_protected_surface.

Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5096>

v3dv: expose VK_KHR_maintenance1

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7364>

v3dv: implement vkTrimCommandPool

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7364>

v3dv: update assertion to match VK_KHR_maintenance1 semantics

Fixes crashes in:
dEQP-VK.api.copy_and_blit.core.image_to_image.3d_images.*

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7364>

v3dv: fix base slice selection for copies involving 3D images

For 3D images we should take the slice to copy from or to, from
the Z coordinate of the corresponding offset, not the base array
layer.

Fixes VK_KHR_maintenance1 tests:
dEQP-VK.api.copy_and_blit.core.image_to_image.3d_images.3d_to_2d_by_slices
dEQP-VK.api.copy_and_blit.core.image_to_image.3d_images.2d_to_3d_by_layers

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7364>

v3dv: add image view debug checks for VK_KHR_maintenance1

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7364>

radv,aco: optimize computing the sample mask for per-sample shading

I don't know why these values were introduced for but it seems like
we can optimize this by just doing:

gl_SampleMaskIn[0] = (SampleCoverage & (1 << gl_SampleID))

AMDGPU-PRO and AMDVLK apply the same formula to compute the
sample mask when per-sample shading is enabled.

No fossils-db changes.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7377>

radv,aco: adjust the sample mask only if per-sample shading is enabled

When per-sample shading isn't enabled, we can just load the
samplemask from the hardware which is always the coverage of
the entire pixel/fragment.

fossilds-db (VEGA10):
Totals from 131 (0.10% of 136546) affected shaders:
SGPRs: 5056 -> 5048 (-0.16%)
VGPRs: 2600 -> 2372 (-8.77%)
CodeSize: 115788 -> 112560 (-2.79%)
MaxWaves: 1266 -> 1274 (+0.63%)
Instrs: 20620 -> 20071 (-2.66%)
Cycles: 82416 -> 80220 (-2.66%)
VMEM: 51567 -> 35532 (-31.10%); split: +0.24%, -31.34%
SMEM: 8952 -> 8258 (-7.75%); split: +0.11%, -7.86%
SClause: 1223 -> 1199 (-1.96%); split: -2.62%, +0.65%
Copies: 1247 -> 1124 (-9.86%); split: -10.18%, +0.32%
PreVGPRs: 2112 -> 1981 (-6.20%)

Helps Britannia, Shadow of the Tomb Raider, Warhammer II and Control.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7377>

glsl: drop NMS OpenGL workarounds

No Mans Sky dropped its OpenGL backend on April 16, 2019 in favour
of its Vulkan backend. So here we drop the old OpenGL workarounds.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7362>

panfrost: Add missing alpha-first special formats

Not sure if these come up with OpenGL in practice.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7352>

panfrost: Fix BGR233 component order

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7352>