intel/blorp: Use designated initializers to set up VERTEX_ELEMENTS

We also add a slot variable and use it as an iterator.  This will make
it much easier to conditionally put something between the header and the
vertex position.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Rename emit_viewport_state to emit_cc_viewport

The real point of this packet is that it sets up CC_VIEWPORT so that
name is a bit better.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Make the common genX_blorp_exec code gen4-safe

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Re-arrange blorp_genX_exec.h

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Don't use ffma directly

It isn't supported prior to gen6 and, on gen6+, NIR will fuse the fmul
and fadd into an ffma automatically for us anyway.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Delete isl_to_gen_ds_surfype

It's no longer used.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Pull the pipeline bits of blorp_exec into a helper

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp/blit: Add support for normalized coordinates

Gen5 and earlier can't do non-normalized coordinates so we need to
compensate in the shader.  Fortunately, it's pretty easy plumb through.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Move clip program compilation to the compiler

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Move SF compilation to the compiler

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/clip: Make brw_clip_prog_key::interp_mode an array

Having it be a pointer means that we end up caching clip programs based
on a pointer to wm_prog_data rather than the actual interpolation modes.
We've been caching one clip program per FS ever since 91d61fbf7cb61a44a
where Timothy rewrote brw_setup_vue_interpolation().

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/sf: make brw_sf_prog_key::interp_mode an array

Having it be a pointer means that we end up caching clip programs based
on a pointer to wm_prog_data rather than the actual interpolation modes.
We've been caching one clip program per FS ever since 91d61fbf7cb61a44a
where Timothy rewrote brw_setup_vue_interpolation().

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/compiler: Make brw_disasm take const assembly

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/decoder: Handle the BLT ring in gen_group_get_length

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

intel/decoder: Handle gen4 VF_STATISTICS and PIPELINE_SELECT

These need special handling because they have no "DWord Length"
parameter and they have an unusual bias of 1.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

intel/genxml: Rename 3DSTATE_AA_LINE_PARAMS on gen5

All of the other gens use "PARAMETERS".

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Use the right subtype for VF_STATISTICS on gen4

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Iron Lake doesn't support non-normalized sampler coordinates

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Add SAMPLER_STATE to gen 4.5

Somehow this got missed.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Rename the CC_VIEWPORT pointer on gen4-5

It isn't a pointer to "color calc state", that's the packet it's in.
It's a pointer to the CC viewport state.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Sampler state is a pointer on gen4-5

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Suffix KSP0 fields on Iron Lake

Iron Lake introduced the multiple KSP thing and so you have KSP0-3.
However, the genxml didn't have an index on the first "Kernel Start
Pointer" or "GRF Register Count".  Add one to match gen6+.  While we're
here, we drop the brackets from the other "GRF Register Count" fields.

Reviewed-by: Matt Turner <mattst88@gmail.com>

intel/genxml: Make a bunch of things offsets on gen4-5

Most things on gen4-5 are addresses because we don't have dynamic state
base address and we don't have instruction state base on gen4.  However,
whoever converted things to addresses got a little over-excited and
converted too much.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/isl: Add gen4_filter_tiling

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/isl: Add support for setting component write disables

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/isl: Add support for gen4 cube maps to get_image_offset_sa

Gen4 cube maps are a 2-D surface with ISL_DIM_LAYOUT_GEN4_3D which is a
bit weird but accurate none the less.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/isl: Don't request space for stencil/hiz packets unless needed

On Iron Lake, the packets exist but we never emit them so there's no
need for us to ask the driver to make batch space for them.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/blorp: Properly handle mt->first_level

The guts of blorp and ISL don't understand i965's partial miptrees.
Instead, we need to subtract off first_level before we hand anything off
to blorp.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/miptree: Take first_level into account when converting to ISL

ISL doesn't have a concept of a partial miptree.  Instead, we need to
subtract off first_level.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Move the gen7 stencil format workaround to blorp_blit

It's not needed for blorp_copy because it already overrides formats.
It's also not needed for blorp_clear because it clears stencil as
stencil.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965: Use blorp_copy for doing r8 stencil updates on HSW

The blorp_copy entrypoint is designed for doing memcpy like operations
which is what we need to do here while blorp_blit is for handling format
conversion and scaling.  Using blorp_copy is much simpler and prevents
us from getting formats wrong.  While we're here, we get rid of the
layers_per_blit thing since stencil always uses interleaved MSAA.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i965/blorp: Do and end-of-pipe sync on both sides of fast-clear ops

We've discovered in the Vulkan driver that simply doing the end-of-pipe
sync afterwards is insufficient.  The specific requirement stated in the
PRM is that you have to do one every time you transition between the
tree modes of "clear", "render", and "resolve".  This is GL, so we could
track it but any attempt to do so would most likely get it wrong.  For
now, it's easier to just assume that every fast-clear op is an island
and do the sync both before and after.

This also removes the unneeded flush and stall after slow-clear
operations.

Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Cc: "17.0 17.1" <mesa-stable@lists.freedesktop.org>

amd/common: add missing libdrm include path

Fixes: de9dd4f9f1bb5984c554 ("ac/radeonsi: move struct radeon_info to ac_gpu_info.h")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>

docs: small release calendar fixes

Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

Revert "amd/common: add vcn dec ip info query"

This reverts commit 524d4fff9e0e58e812e71bc5ba2c92ffa73bee16.

This commit breaks amdgpu on kernels with no DEC IP support.

Caught by the airlied CI system.

Revert "amd/common: set vcn dec as hw decode as well"

This reverts commit 50d322be2f30bb935ee52b3b3649785975771110.

A previous patch breaks amdgpu on non-vcn decode systems,
but have to revert this first.

util: remove unneeded Android ifdef from ralloc.c

SIZE_MAX has been defined in stdint.h on Android since 2013, so this ifdef
is no longer needed.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>

nouveau: drop Android 4.4 and earlier support

Support for Android 4.4 and earlier has already been removed from mesa.
Remove this remaining piece from nouveau, too.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>

i965: use mmap64 for Android

Simplify the handling of mmap for Android by using mmap64 instead. mmap64
may have not existed for Android when this was written, but it's been
around since 2013.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>

gallium/os: use mmap64 for Android

Simplify the handling of mmap for Android by using mmap64 instead. mmap64
may have not existed for Android when this was written, but it's been
around since 2013.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>

Android: generate an error if building on Android 4.4 or earlier

Since commit 7a5b5f52260 ("Android: drop Android 4.4 (KitKat) support"),
Android 4.4 or earlier is no longer supported, so exit with an error if we
try building on it.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>

st/wgl: whitespace, formatting fixes in stw_device.c

Trivial.

glsl: Fix g++ initializer order warning

Fixes this warning:
In file included from ../../../src/compiler/glsl/ir.cpp:25:0:
../../../src/compiler/glsl/ir.h: In constructor 'ir_swizzle::ir_swizzle(ir_rvalue*, ir_swizzle_mask)':
../../../src/compiler/glsl/ir.h:1955:20: warning: 'ir_swizzle::mask' will be initialized after [-Wreorder]
    ir_swizzle_mask mask;
                    ^
../../../src/compiler/glsl/ir.h:1954:15: warning:   'ir_rvalue* ir_swizzle::val' [-Wreorder]
    ir_rvalue *val;
               ^
../../../src/compiler/glsl/ir.cpp:1592:1: warning:   when initialized here [-Wreorder]
 ir_swizzle::ir_swizzle(ir_rvalue *val, ir_swizzle_mask mask)
 ^

Reviewed-by: Matt Turner <mattst88@gmail.com>

radeonsi: enable vcn decode

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

winsys/amdgpu: add vcn dec cs support

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amd/common: set vcn dec as hw decode as well

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

amd/common: add vcn dec ip info query

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon: rename has_uvd info to has_hw_decode

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message for mpeg4 codec

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message for mpeg2 codec

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message for vc1 codec

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message for hevc codec

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message decode for avc codec

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message feedback

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message destroy

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add decode message create

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add common decode part

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/winsys: add vcn dec ring type

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/winsys: add uvd enc ring type

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

radeon/vcn: add vcn decode interface

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

configure.ac: update libdrm amdgpu version requirement to 2.4.81

VCN decode has a new interface, and that depends on the latest libdrm

Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

docs: update calendar, add news item and link release notes for 17.1.1

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

docs: add sha256 checksums for 17.1.1

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
(cherry picked from commit 092c485b8ed127a15f52434857c37e8021775775)

docs: add release notes for 17.1.1

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
(cherry picked from commit ca0a148a4d09d850f63c82011ac49a311e43e502)

mesa: remove unrequired double calc

type_size() will already handle this correctly for us.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

mesa: remove redundant modulus operation

The if check above means we can only get here if size is less than 4.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

svga: init svga_screen::swc_mutex with mtx_recursive

If the SVGA3D_BindGBSurface() call in svga_buffer_hw_storage_unmap()
fails, we'll flush and that might involve unmapping other buffers.
That leads to a recursive lock on svga_screen::swc_mutex and causes
a deadlock.  Fix this by initializing the mutex with mtx_recursive.

Note that this only happened on Linux, not Windows.  On Windows, the
mutex functions are implemented with Win32 critical sections which
support recursive locking.

Also add a comment about this.

Fixes VMware bug 1831549 (Unigine Tropics demo freeze on Linux).

Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Neha Bhende<bhenden@vmware.com>

svga: move logging initialization code into new function

Plus a few other minor clean-ups.

Reviewed-by: Sinclair Yeh <syeh@vmware.com>

svga: init local vars to silence uninitialized use warnings

Reviewed-by: Sinclair Yeh <syeh@vmware.com>

svga: log the process command line to the vmware.log file

This is useful for Piglit when thousands of tests are run and we want
to determine which test triggered a device error.

v2: only log command line info if the new SVGA_EXTRA_LOGGING env var is set

Reviewed-by: Charmaine Lee <charmainel@vmware.com>

svga: Limit svga message capability to newer compilers

The assembly code used by the SVGA message feature doesn't
build properly with older compilers, so limit it to only
gcc 5.3.0 and newer.

Also modified the stubs to avoid "unused variable" warnings.

Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

svga: Fix MSVC build.

This let us compile the code with MSVC, but it no-ops the log function.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

svga: Add the ability to log messages to vmware.log on the host.

For now this capability only exists in the SVGA driver but
can be exported later if other modules, e.g. winsys, wants
to use it for logging.

Reviewed-by: Brian Paul <brianp@vmware.com>

Revert "gallium: remove unused PIPE_CC_GCC_VERSION"

This reverts commit e60928f4c4bd4484821d83f2b16a910ea9f5f9d9.

PIPE_CC_GCC_VERSION is used by some of our in-house code which hasn't
been upstreamed yet.

aubinator: report error on unknown device id

Since we're going to stop aubinator without a valid device id, better
report an error. This also silences a Coverity warning.

CID: 1405004
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

aubinator: be consistent on exit code

We're using both exit(1) & exit(EXIT_FAILURE), settle for one, same
for success.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

aubinator: fix double free
1;4601;0c
Free previously allocated filename outside the for loop.

CID: 1405014
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

winsys/amdgpu: align VA allocations to fragment size v2

BOs larger than the minimum fragment size should have their VA
alignet to at least the fragment size for optimal performance.

v2: drop unused leftover from initial implementation

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

tgsi: remove unused tgsi_is_passthrough_shader()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Elie Tournier <elie.tournier@collabora.com>

configure.ac: rephrase 'GLX w/o X11' error message

Signed-off-by: Eric Engestrom <eric@engestrom.ch>

anv: Require vertex buffers to come from a 32-bit heap

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Advertise both 32-bit and 48-bit heaps when we have enough memory

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Refactor memory type setup

This makes us walk over the heaps one at a time and add the types for
LLC and !LLC to each heap.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Make supports_48bit_addresses a heap property

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Stop setting BO flags in bo_init_new

The idea behind doing this was to make it easier to set various flags.
However, we have enough custom flag settings floating around the driver
that this is more of a nuisance than a help.  This commit has the
following functional changes:

 1) The workaround_bo created in anv_CreateDevice loses both flags.
    This shouldn't matter because it's very small and entirely internal
    to the driver.

 2) The bo created in anv_CreateDmaBufImageINTEL loses the
    EXEC_OBJECT_ASYNC flag.  In retrospect, it never should have gotten
    EXEC_OBJECT_ASYNC in the first place.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Set image memory types based on the type count

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Add valid_bufer_usage to the memory type metadata

Instead of returning valid types as just a number, we now walk the list
and check the buffer's usage against the usage flags we store in the new
anv_memory_type structure.  Currently, valid_buffer_usage == ~0.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Determine the type of mapping based on type metadata

Before, we were just comparing the type index to 0.  Now we actually
look the type up in the table and check its properties to determine what
kind of mapping we want to do.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Set up memory types and heaps during physical device init

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Predicate 48bit support on gen >= 8

This doesn't matter right now since it only affects whether or not we
set the kernel bit but, if we ever do anything else based on it, we'll
want it to be correct per-gen.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv/image: Get rid of the memset(aux, 0, sizeof(aux)) hack

Up until now, we've been memsetting the auxiliary surface to 0 at
BindImageMemory time to ensure that it is properly initialized.
However, this isn't correct because apps are allowed to freely alias
memory between different images and buffers so long as they properly
track whether or not a particular image is valid and, if it isn't,
transition from UNINITIALIZED to something else before using it.  We
now implement those transitions so we can drop the hack.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Handle transitioning depth from UNDEFINED to other layouts

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

anv: Handle color layout transitions from the UNINITIALIZED layout

This causes dEQP-VK.api.copy_and_blit.resolve_image.partial.* to start
failing due to test bugs.  See CL 1031 for a test fix.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Cc: "17.1" <mesa-stable@lists.freedesktop.org>

st/nine: Fix a regression and syntax cleanup

A few cleanups and in particular initializing properly
the new pipe_draw_info fields.
This should fix the regression caused by
330d0607ed60fd3edca192e54b4246310f06652f

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101088

Signed-off-by: Axel Davy <axel.davy@ens.fr>
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>

mesa: Remove GL_APPLE_vertex_array_object stubs

Mark the functions 'exec="skip"' in the XML instead.  libGL will still
have the functions, but the driver won't try to use them.  I verified
that this commit works with piglit's 'object-namespace-pollution glClear
vertex-array' on x64 with a driver built from mesa-12.0.3 tag.

In fairness, this test also works with a libGL built from 7927d03.  I
believe it continues to work because on non-Windows platforms we
generate some extra, dummy dispatch functions that can be used when a
driver requests a function unknown to libGL.  This was done to provide
some "forward" compatibility with drivers that need more functions.
This doesn't work on Windows because the Windows calling convention is
for the callee to clean up the stack.  That's the theory anyway.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

gallium/radeon: pipe AMDGPU_INFO_NUM_VRAM_CPU_PAGE_FAULTS into gallium HUD

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

freedreno/ir3: switch to NIR by default

Now that we lower vars to regs, we no longer regress for anything that
does complex dereferences.  (With tgsi, derefers are already lowered
before tgsi_to_nir, but not with glsl_to_nir.)  In fact it actually
fixes a few things to bypass tgsi.

So make NIR the default (finally!)

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: lower arrays to regs

Instead of using load/store_var intrinsics, which can have complex
derefs in the case of multi-dimensional arrays, lower these to regs
and handle the direct/indirect loads in get_src() and stores in
put_dst().

This should let us switch to using nir by default.

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: add put_dst()

Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno/ir3: code-motion

Signed-off-by: Rob Clark <robdclark@gmail.com>