ac/nir: Implement derefs for integer gather4 lowering.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

ac/nir: Support deref instructions in tex instructions.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

ac/nir: Support deref instructions in get_sampler_desc.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

ac/nir: Implement the deref instr for shared memory.

v2: Store the result in ctx->ssa_defs.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir,spirv: Rework function calls

This commit completely reworks function calls in NIR.  Instead of having
a set of variables for the parameters and return value, nir_call_instr
now has simply has a number of sources which get mapped to load_param
intrinsics inside the functions.  It's up to the client API to build an
ABI on top of that.  In SPIR-V, out parameters are handled by passing
the result of a deref through as an SSA value and storing to it.

This virtue of this approach can be seen by how much it allows us to
delete from core NIR.  In particular, nir_inline_functions gets halved
and goes from a fairly difficult pass to understand in detail to almost
trivial.  It also simplifies spirv_to_nir somewhat because NIR functions
never were a good fit for SPIR-V.

Unfortunately, there is no good way to do this without a mega-commit.
Core NIR and SPIR-V have to be changed at the same time.  This also
requires changes to anv and radv because nir_inline_functions couldn't
handle deref instructions before this change and can't work without them
after this change.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv/cfg: Make the builder fully capable for both walks

We were only initializing vtn_builder::func for the pre-walk where we
build the CFG.  We were only initializing the nir_builder for the later
walk through the instructions even though were were setting b->cursor
for the pre-walk.  Let's set both both places so that everything is
consistent.  This useful because we handle OpFunctionParameter in the
pre-walk and we're going to need to be able to emit instructions.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Record the type of functions

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Update vtn_pointer_to/from_ssa to handle deref pointers

Now that pointers can be derefs and derefs just produce SSA values, we
can convert any pointer to/from SSA.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Allow pointers to have a deref at the base

Previously, pointers fell into two categories: index/offset for UBOs,
SSBOs, etc. and var + access chain for logical pointers.  This commit
adds another logical pointer mode that's deref + access chain.

It's tempting to think that we can just replace variable-based pointers
with deref-based or at least replace the access chain with a deref
chain.  Unfortunately, there are a few sticky bits that prevent this:

 1) We can't return deref-based pointers from OpVariable because those
    opcodes may come outside of a function so there's no place to emit
    the deref instructions.

 2) We can't always use variable-based pointers because we may not
    always know the variable.  (We do now, but he upcoming function
    rework will take that option away.)

 3) We also can't replace the access chain struct with a deref.  Due to
    the re-ordering we do in order to handle loop continues, the derefs
    we would emit as part of OpAccessChain may not dominate their uses.
    We normally fix this up with nir_repair_ssa but that generates phi
    nodes which we don't want in the middle of our deref chains.

All in all, we have no real better option than to support partial access
chains while also re-emitting the deref instructions on the spot.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Clean up vtn_pointer_to_offset

Now that push constants are using on-the-fly offsets, we no longer need
to handle access chains in vtn_pointer_to_offset.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Make push constants an offset-based pointer

Push constants have been a weird edge-case for a while in that they have
explitic offsets but we've been internally building access chains for
them.  This mostly works but it means that passing pointers to push
constants through as function arguments is broken.  The easy thing to do
for now is to just treat them like UBOs or SSBOs only without a block
index.  This does loose a bit of information since we no longer have an
accurate access range and any indirect access will look like it could
read the whole block.  Unfortunately, there's not much we can do about
that.  Once NIR derefs get a bit more powerful, we can plumb these
through as derefs and be able to reason about them again.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Use NIR per-member splitting

Before, we were doing structure splitting in spirv_to_nir.
Unfortunately, this doesn't really work when you think about passing
struct pointers into functions.  Doing it later in NIR is a much better
plan.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/spirv: Pass nir_variable_data into apply_var_decoration

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add a concept of per-member structs and a lowering pass

This adds a concept of "members" to a variable with an interface type.
It allows you to specify the full variable data for each member of the
interface instead of once for the variable.  We also add a lowering pass
to lower those variables to a sequence of variables and rewrite all the
derefs accordingly.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

spirv: Use deref instructions for most variables

The only thing still using old-school drefs are function calls.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

st/nir: Move lower_deref_instrs later

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: Move nir_lower_deref_instrs to right before locals_to_regs

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/lower_tex: Always copy deref and offset sources

This should make nir_lower_tex properly handle deref instructions as
well as make it more correct when texture arrays are used and it's
called after lowering samplers to binding table indices.

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/nir: Fixup deref modes after lowering patch vertices

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel,ir3: Disable nir_opt_copy_prop_vars

This pass doesn't handle deref instructions yet.  Making it handle both
legacy derefs and deref instructions would be painful.  Since it's not
important for correctness, just disable it for now.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in opt_undef

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Consider deref instructions in opt_peephole_select

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Consider deref instructions in lower_phis_to_scalar

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_drawpixels

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_clamp_color_outputs

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_alpha_test

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in loop_analyze

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in remove_unused_varyings

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_pos_center

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_wpos_ytransform

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_atomics

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_io

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in gather_info

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in propagate_invariant

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_clip_cull

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_system_values

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_indirect_derefs

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_vars_to_ssa

Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in split_var_copies

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_var_copies

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add a deref path helper struct

This commit introduces a new nir_deref.h header for helpers that are
less common and really only needed by a few heavy-duty passes.  In this
header is a new struct for representing a full deref path which can be
walked in either direction.

v2 (Jason Ekstrand):
 - Assert that deref != NULL (Caio)
 - Fill _short_path with 0xdeadbeef in debug builds when not used (Caio)
 - Make nir_deref_path a typedef (Rob)

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_io_to_temporaries

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in lower_global_vars_to_local

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add a pass for fixing deref modes

This will be needed by anything which changes variable modes without
rewriting derefs.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Support deref instructions in remove_dead_variables

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

ttn: convert to deref instructions

Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

prog/nir: Use deref instructions for params

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl/nir: Use deref instructions instead of dref chains

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl/nir: Only claim to handle intrinsic functions

Non-intrinsic function handling has never actually been tested and
probably doesn't work.  Just get rid of it for now.  We can always add
it back in later if it's useful.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: add deref lowering sanity checking

This will be removed at the end of the transition, but add some tracking
plus asserts to help ensure that lowering passes are called at the
correct point (pre or post deref instruction lowering) as passes are
converted and the point where lower_deref_instrs() is called is moved.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

anv,i965,radv,st,ir3: Call nir_lower_deref_instrs

This inserts a call to nir_lower_deref_instrs at every call site of
glsl_to_nir, spirv_to_nir, and prog_to_nir.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/deref: Add some deref cleanup functions

Sometimes it's useful for a pass to be able to clean up its own derefs
instead of waiting for DCE.  This little helper makes it very easy.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add helpers for working with deref instructions

This commit adds a pass for lowering deref instructions to deref chains
as well as some smaller helpers to ease the transition.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add deref sources to texture instructions

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add _deref versions of all of the _var intrinsics

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/builder: Add deref building helpers

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir: Add a deref instruction type

This commit adds a new instruction type to NIR for handling derefs.
Nothing uses it yet but this adds the data structure as well as all of
the code to validate, print, clone, and [de]serialize them.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/validate: Rework intrinsic type validation

This moves the switch statement for specific intrinsics above source and
destination validation.  We also rework the source and destination
validation to use different bit_size values for each source and/or
destination.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nv50/ir: only avoid spilling constrained def if a mov is added

fix spilling regression introduced by 5428066f5e

this is just a minor mistake done while moving the code out into a new
function. The function contained a loop which might have been terminated
earlier and skipped setting noSpill to 1. After the refactoring it was always
set.

Fixes: 5428066f5e1ef5ea6ae04c84019f270023cfc6aa
("nv50/ir: make a copy of tex src if it's referenced multiple times")
Signed-off-by: Karol Herbst <kherbst@redhat.com>

meson: Fix typo that breaks -Dgalium-xvmc=false

_xmvc -> _xvmc. Sigh

Fixes: a6943bb4ce99ea7af7222552f1326fb80823d0b2
       ("meson: Fix auto option for xvmc")
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Tested-by: Clayton Craft <clayton.a.craft@intel.com>

meson: Fix auto option for va

The same as the previous two patches, but for the libva state tracker.

Fixes: 724916c8a84b5bba8f880f17da936585d52c97b6
       ("meson: dedup gallium-xvmc logic")
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

meson: Fix auto option for xvmc

This fixes the same problem as the previous patch did for vdpau, but for
xvmc.

Fixes: 724916c8a84b5bba8f880f17da936585d52c97b6
       ("meson: dedup gallium-xvmc logic")
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

meson: Correct behavior of vdpau=auto

Currently if vdpau is set to auto, it will be disabled only in cases
where gallium is disabled or the host OS is not supported (mac, haiku,
windows). However on (for example) Linux if libvdpau is not installed
then the build will error because of the unmet dependency. This corrects
auto to do the right thing, and not error if libvdpau is not installed.

Fixes: 992af0a4b8224bdb4809e01c2f00d2f32546aee5
       ("meson: dedup gallium-vdpau logic")
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

radv: always check the return error when submitting a CS

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: check the return values of radv_signal_fence()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: change the returned error in radv_signal_fence()

From my point of view, when we aren't able to submit a CS
something terribly wrong happens and we are most likely
going to lost the device.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

freedreno: a2xx: fix clear color

the format of the CLEAR_COLOR register doesn't depend on the target format
this fixes clear color when rendering to 32-bit RGBA and 16-bit targets

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: a2xx: fix crash when freeing context

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: a2xx: fix crash on first clear

blend can be NULL, so check for that

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: add a20x

this patch adds support for a20x, which has some differences with a220:
-no VGT_MAX_VTX_INDX register
-no CLEAR_COLOR register
-set RB_BC_CONTROL in restore (hangs without)
-different CP_DRAW_INDX format

tested with kmscube and glmark2 scenes, on par with a220

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

freedreno: a2xx: increase size of the offset field in instr_fetch_vtx_t

The offset field is 22 bit large.
11 bits are necessary because MaxVertexAttribRelativeOffset = 2047

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Signed-off-by: Rob Clark <robdclark@gmail.com>

v3d: Don't forget to initialize the buffer offset of a new winsys handle.

v3d, vc4: Disable valgrind checking of CLE inputs when NDEBUG is set.

For a meson -Db_ndebug=true release build on x86_64, reduces text size of
libv3d.a from 53.0k to 51.6k.  Inspired by 0d5329d626e3 ("anv: Disable
__gen_validate_value if NDEBUG is set.")

mesa: fix glGetInteger64v for arrays of integers

Cc: 18.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

ac/surface: disallow rotated micro tile mode

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radeonsi: fix occlusion queries with 16x AA without FBO attachments on Stoney

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radeonsi: handle non-clearable DCC buffers as MSAA resolve dst

This is reproducible on Stoney, but other chips may be affected too.

Cc 18.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radeonsi: disable DCC MSAA for 128bpp formats on Stoney

Cc: 18.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

docs: update freedreno features

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

mesa: fix GLES 3.1 version calculation

All of ARB_gpu_shader5 is most certainly not required for GLES 3.1
(most of it is in OES_gpu_shader5 on top of GLES 3.1).

Some of what is required from ARB_gpu_shader5 is provided by
ARB_texture_gather, so check for that.  The remaining subset of
ARB_gpu_shader5 doesn't have individual extensions to check for,
but I guess it is unlikely that some driver has all of these
extensions but not, say, integer bitfield manipulation.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

freedreno/a5xx: MSAA

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

freedreno: update generated headers

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

freedreno/ir3: txf_ms support

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

freedreno/a5xx: fix gpu hangs with large compute shaders

Similar to the combined limit for VS+FS, there is an upper limit for
shader size to run from internel memory.

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

freedreno/ir3: fix base_vertex

Fixes: c366f422f0a nir: Offset vertex_id by first_vertex instead of base_vertex
Signed-off-by: Rob Clark <robdclark@gmail.com>

i965: Link uniforms of SPIR-V programs using the NIR linker

v2: nir_link_uniforms renamed to gl_nir_link_uniforms

Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

i965: Setup glsl uniforms by index rather than name matching

Previously when setting up a uniform it would try to walk the uniform
storage slots and find one that matches the name of the given
variable. However, each variable already has a location which is an
index into the UniformStorage array so we can just directly jump to
the right slot. Some of the variables take up more than one slot so we
still need to calculate how many it uses.

The main reason to do this is to support ARB_gl_spirv because in that
case the uniforms don’t have names so the previous approach won’t
work.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

i965: account for NIR uniforms without name

Right now, the BRW linker code assumes nir_variable::name is always
non-NULL, but thanks to ARB_gl_spirv we will soon be linking SPIR-V
programs, and those explicitly require matching uniforms by location.
The name is just a debug hint.

Instead of checking for the name this patch makes it check for
var->num_state_slots on the assumption that everything that had an
internal name also had some state slots. This seems likely because the
two code paths that are taken when the name begins with "gl_" already
have an assert that var->state_slots is not NULL.

v2: simplified, most of it moved to glsl/nir/spirv (Neil Roberts)
v3: check for num_state_slots instead of the name. This is needed
    because we do actually have nameless builtins with SPIR-V such as
    PatchVerticesIn and we want them to hit the
    _mesa_add_state_reference code path (Neil Roberts)

Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Neil Roberts <nroberts@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

i965: Update TexturesUsed after linking the shaders

Otherwise if the shader is SPIR-V then SamplerUsed won’t have been
initialised yet so it will end up thinking no textures are used. This
was causing a crash later on if nothing causes it to regenerate
TexturesUsed before the next render.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

i965: Build SPIR-V programs' resource list using NIR

v2: tweak after nir_linker.h being renamed to gl_nir_linker.h

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

nir/linker: Add nir_build_program_resource_list()

This function is equivalent to the linker.cpp
build_program_resource_list() but will extract the resources from NIR
shaders instead.

For now, only uniforms and program inputs are implemented.

v2: move from compiler/nir to compiler/glsl (Timothy Arceri)

v3: remove support for inputs, that is still WIP (spotted by Timothy
    Arceri)

Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

compiler/link: move add_program_resource to linker_util

So it could be used by the GLSL and NIR linker.

v2: (Timothy Arceri)
   * Moved from compiler to compiler/glsl
   * Method renamed to link_util_add_program_resource

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

nir/linker: Set the uniform initial values

This is based on link_uniform_initializers.cpp.

v2: move from compiler/nir to compiler/glsl (Timothy Arceri)

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

nir/linker: Add gl_nir_link_uniforms()

This function will be the entry point for linking the uniforms from
the nir_shader objects associated with the gl_linked_shaders of a
program.

This patch includes initial support for linking uniforms from NIR
shaders. It is tailored for the ARB_gl_spirv needs, and it is far from
complete, but it should handle most cases of uniforms, array
uniforms, structs, samplers and images.

There are some FIXMEs related to specific features that will be
implemented in following patches, like atomic counters, UBOs and
SSBOs.

Also, note that ARB_gl_spirv makes mandatory explicit location for
normal uniforms, so this code only handles uniforms with explicit
location. But there are cases, like uniform atomic counters, that
doesn't have a location from the OpenGL point of view (they have a
binding), but that Mesa assign internally a location. That will be
handled on following patches.

A nir_linker.h file is also added. More NIR-linking related API will
be added in subsequent patches and those will include stuff from Mesa,
so reusing nir.h didn't seem a good idea.

v2: move from compiler/nir to compiler/glsl (Timothy Arceri)
v3: sets var->driver.location if the uniform was found from a previous
    stage (Neil Roberts).

Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Neil Roberts <nroberts@igalia.com
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

compiler/link: add linker_util.h, move linker_error/warning to it

Linker utilities common to the GLSL IR and NIR linker (the latter to
be used for ARB_gl_spirv).

We need to move it to a new header as the NIR linker doesn't need to
know about ir_variable, and others, included at linker.h.

v2: move from src/compiler to src/compiler/glsl (Timothy Arceri)

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

spirv: Set nir_variable->explicit_binding

When SpvDecorationBinding is encountered in the SPIR-V source it now
sets explicit_binding on the nir_variable. This will be used to
determine whether to initialise sampler and image uniforms with the
binding value.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

spirv: Get rid of vtn_variable_mode_image/sampler

vtn_variable_mode_image and _sampler are instead replaced with
vtn_variable_mode_uniform which encompasses both of them. In the few
places where it was neccessary to distinguish between the two, the
GLSL type of the pointer is used instead.

The main reason to do this is that on OpenGL it is permitted to put
images and samplers into structs and declare a uniform with them. That
means that variables can now have a mix of uniform, sampler and image
modes so picking a single one of those modes for a variable no longer
makes sense.

This fixes OpLoad on a sampler within a struct which was previously
using the variable mode to determine whether it was a sampler or not.
The type of the variable is a struct so it was not being considered to
be uniform mode even though the member being loaded should be sampler
mode.

The previous code appeared to be using var->interface_type as a place
to store the type of the variable without the enclosing array for
images and samplers. I guess this worked because opaque types can not
appear in interfaces so the interface_type is sort of unused. This
patch removes the overloading of var->interface_type and any places
that needed the type without the array can now just deduce it from
var->type.

v2: squash in this patch the changes to anv/nir (Timothy)

Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Neil Roberts <nroberts@igalia.com
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

spirv: translate default-block uniforms

They are supported by SPIR-V for ARB_gl_spirv.

v2 (changes on top of Nicolai's original patch):
   * Handle UniformConstant storage class for uniforms other than
     samplers and images. (Eduardo Lima)
   * Handle location decoration also for samplers and images. (Eduardo
     Lima)
   * Rebase update (spirv_to_nir options added, logging changes, and
     others) (Alejandro Piñeiro)

Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

nir/types: Add a utility wrapper to glsl_type::sampler_index()

I think it is more accurate to call it a sampler target (?).

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

nir/types: Add a glsl_get_component_slots() utility

It is basically a wrapper around glsl_type::component_slots().

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>