Previously, patch constant functions only accepted OutputPatch. This
adds InputPatch support, via a pseudo-builtin variable type, so that
the patch can be tracked clear through from the qualifier.
In the hull shader, the PCF output does not participate in an argument list,
so has no defined ordering. It is always put at the end of the linkage. That
means the DS input reading PCF data must be be at the end of the DS linkage
as well, no matter where it may appear in the argument list. This change
makes sure that happens.
The detection is by looking for arguments that contain tessellation factor
builtins, even as a struct member. The whole struct is taken as the PCF output
if any members are so qualified.
The SPIR-V generator had assumed tessellation modes such as
primitive type and vertex order would only appear in tess eval
(domain) shaders. SPIR-V allows either, and HLSL allows and
possibly requires them to be in the hull shader.
This change:
1. Passes them through for either tessellation stage, and,
2. Does not set up defaults in the domain stage for HLSl compilation,
to avoid conflicting definitions.
Unknown how extensive the semantics need to be yet. Need real
feedback from workloads. This is just done as part of unifying it
with the class/struct namespaces and grammar productions.
This PR emulates per control point inputs to patch constant functions.
Without either an extension to look across SIMD lanes or a dedicated
stage, the emulation must use separate invocations of the wrapped
entry point to obtain the per control point values. This is provided
since shaders are wanting this functionality now, but such an extension
is not yet available.
Entry point arguments qualified as an invocation ID are replaced by the
current control point number when calling the wrapped entry point. There
is no particular optimization for the case of the entry point not having
such an input but the PCF still accepting ctrl pt frequency data. It'll
work, but anyway makes no so much sense.
The wrapped entry point must return the per control point data by value.
At this time it is not supported as an output parameter.
This PR adds the ability to pass structuredbuffer types by reference
as function parameters.
It also changes the representation of structuredbuffers from anonymous
blocks with named members, to named blocks with pseudonymous members.
That should not be an externally visible change.
This is a partial implemention of structurebuffers supporting:
* structured buffer types of:
* StructuredBuffer
* RWStructuredBuffer
* ByteAddressBuffer
* RWByteAddressBuffer
* Atomic operations on RWByteAddressBuffer
* Load/Load[234], Store/Store[234], GetDimensions methods (where allowed by type)
* globallycoherent flag
But NOT yet supporting:
* AppendStructuredBuffer / ConsumeStructuredBuffer types
* IncrementCounter/DecrementCounter methods
Please note: the stride returned by GetDimensions is as calculated by glslang for std430,
and may not match other environments in all cases.
This obsoletes WIP PR #704, which was built on the pre entry point wrapping master. New version
here uses entry point wrapping.
This is a limited implementation of tessellation shaders. In particular, the following are not functional,
and will be added as separate stages to reduce the size of each PR.
* patchconstantfunctions accepting per-control-point input values, such as
const OutputPatch <hs_out_t, 3> cpv are not implemented.
* patchconstantfunctions whose signature requires an aggregate input type such as
a structure containing builtin variables. Code to synthesize such calls is not
yet present.
These restrictions will be relaxed as soon as possible. Simple cases can compile now: see for example
Test/hulsl.hull.1.tesc - e.g, writing to inner and outer tessellation factors.
PCF invocation is synthesized as an entry point epilogue protected behind a barrier and a test on
invocation ID == 0. If there is an existing invocation ID variable it will be used, otherwise one is
added to the linkage. The PCF and the shader EP interfaces are unioned and builtins appearing in
the PCF but not the EP are also added to the linkage and synthesized as shader inputs.
Parameter matching to (eventually arbitrary) PCF signatures is by builtin variable type. Any user
variables in the PCF signature will result in an error. Overloaded PCF functions will also result in
an error.
[domain()], [partitioning()], [outputtopology()], [outputcontrolpoints()], and [patchconstantfunction()]
attributes to the shader entry point are in place, with the exception of the Pow2 partitioning mode.
Structs are split to remove builtin members to create valid SPIR-V. In this
process, an outer structure array dimension may be propegated onto the
now-removed builtin variables. For example, a mystruct[3].position ->
position[3]. The copy between the split and unsplit forms would handle
this in some cases, but not if the array dimension was at different levels
of aggregate.
It now does this, but may not handle arbitrary composite types. Unclear if
that has any semantic meaning for builtins though.
Previously, a type graph would turn into a type tree. That is,
a deep node that is shared would have multiple copies made.
This is important when creating IO and non-IO versions of deep types.
This needs some render testing, but is destined to be part of master.
This also leads to a variety of other simplifications.
- IO are global symbols, so only need one list of linkage nodes (deferred)
- no longer need parse-context-wide 'inEntryPoint' state, entry-point is localized
- several parts of splitting/flattening are now localized
When copying split types with mixtures of user variables and buitins,
where the builtins are extracted, there is a parallel structures traversal.
The traversal was not obtaining the derefenced types in the array case.
This partially addressess issue #670, for when the matrix swizzle
degenerates to a component or column: m[c], m[c][r] (where HLSL
swaps rows and columns for user's view).
An error message is given for the arbitrary cases not covered.
These cases will work for arbitrary use of l-values.
Future work will handle more arbitrary swizzles, which might
not work as arbitrary l-values.
Reads and write syntax to UAV objects is turned into EOpImageLoad/Store
operations. This translation did not support destination swizzles,
for example, "mybuffer[tc].zyx = 3;", so such statements would fail to
compile. Now they work.
Parial updates are explicitly prohibited.
New test: hlsl.rw.swizzle.frag
This PR adds support for default function parameters in the following cases:
1. Simple constants, such as void fn(int x, float myparam = 3)
2. Expressions that can be const folded, such a ... myparam = sin(some_const)
3. Initializer lists that can be const folded, such as ... float2 myparam = {1,2}
New tests are added: hlsl.params.default.frag and hlsl.params.default.err.frag
(for testing error situations, such as ambiguity or non-const-foldable).
In order to avoid sampler method ambiguity, the hlsl better() lambda now
considers sampler matches. Previously, all sampler types looked identical
since only the basic type of EbtSampler was considered.
HLSL allows type keywords to also be identifiers, so a sequence such as "float half = 3" is
valid, or more bizzarely, something like "float.float = int.uint + bool;"
There are places this is not supported. E.g, it's permitted for struct members, but not struct
names or functions. Also, vector or matrix types such as "float3" are not permitted as
identifiers.
This PR adds that support, as well as support for the "half" type. In production shaders,
this was seen with variables named "half". The PR attempts to support this without breaking
useful grammar errors such as "; expected" at the end of unterminated statements, so it errs
on that side at the possible expense of failing to accept valid constructs containing a type
keyword identifier. If others are discovered, they can be added.
Also, half is now accepted as a valid type, alongside the min*float types.
This commit adds support for copying nested hierarchical types of split
types. E.g, a struct of a struct containing both user and builtin interstage
IO variables.
When copying split types, if any subtree does NOT contain builtin interstage
IO, we can copy the whole subtree with one assignment, which saves a bunch
of AST verbosity for memberwise copies of that subtree.
This adds structure splitting, which among other things will enable GS support where input structs
are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case.
The salient points of this PR are:
* Structure splitting has been changed from "always between stages" to "only into the VS and out of
the PS". It had previously happened between stages because it's not legal to pass a struct
containing a builtin IO variable.
* Structs passed between stages are now split into a struct containing ONLY user types, and a
collection of loose builtin IO variables, if any. The user-part is passed as a normal struct
between stages, which is valid SPIR-V now that the builtin IO is removed.
* Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g,
functions can work unmodified.
* If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS,
the array reference is moved to the split-off loose variable, which is given the array dimension
itself.
When passing things around inside the shader, such as over a function call, the the original type
is used in a sanitized form that removes the builtIn qualifications and makes them temporaries.
This means internal function calls do not have to change. However, the type when returned from
the shader will be member-wise copied from the internal sanitized one to the external type.
The sanitized type is used in variable declarations.
When copying split types and unsplit, if a sub-struct contains only user variables, it is copied
as a single entity to avoid more AST verbosity.
Above strategy arrived at with talks with @johnkslang.
This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to
real world cases.
This PR implements recursive type flattening. For example, an array of structs of other structs
can be flattened to individual member variables at the shader interface.
This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient
for geometry shader arrayed inputs. That will be handled separately with structure splitting,
which is not implemented by this PR. In the meantime, that case is detected and triggers an error.
The recursive flattening extends the following three aspects of single-level flattening:
- Flattening of structures to individual members with names such as "foo[0].samp[1]";
- Turning constant references to the nested composite type into a reference to a particular
flattened member.
- Shadow copies between arrays of flattened members and the nested composite type.
Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR.
Internally, shadow copies are, such as if the type is passed to a function.
Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct
types are flattened. (The latter will change with structure splitting).
One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be
flattened.
Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently
issues an error until type splitting is online).
The process of arriving at the individual member from chained postfix expressions is more complex than
it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
PR #577 addresses most but not all of the intrinsic promotion problems.
This PR resolves all known cases in the remainder.
Interlocked ops need special promotion rules because at the time
of function selection, the first argument has not been converted
to a buffer object. It's just an int or uint, but you don't want
to convert THAT argument, because that implies converting the
buffer object itself. Rather, you can convert other arguments,
but want to stay in the same "family" of functions. E.g, if
the first interlocked arg is a uint, use only the uint family,
never the int family, you can convert the other args as you please.
This PR allows making such opcode and arg specific choices by
passing the op and arg to the convertible lambda. The code in
the new test "hlsl.promote.atomic.frag" would not compile without
this change, but it must compile.
Also, it provides better handling of downconversions (to "worse"
types), which are permitted in HLSL. The existing method of
selecting upconversions is unchanged, but if that doesn't find
any valid ones, then it will allow downconversions. In effect
this always uses an upconversion if there is one.
This PR handles implicit promotions for intrinsics when there is no exact match,
such as for example clamp(int, bool, float). In this case the int and bool will
be promoted to a float, and the clamp(float, float, float) form used.
These promotions can be mixed with shape conversions, e.g, clamp(int, bool2, float2).
Output conversions are handled either via the existing addOutputArgumentConversion
function, which this PR generalizes to handle either aggregates or unaries, or by
intrinsic decomposition. If there are methods or intrinsics to be decomposed,
then decomposition is responsible for any output conversions, which turns out to
happen automatically in all current cases. This can be revisited once inout
conversions are in place.
Some cases of actual ambiguity were fixed in several tests, e.g, spv.register.autoassign.*
Some intrinsics with only uint versions were expanded to signed ints natively, where the
underlying AST and SPIR-V supports that. E.g, countbits. This avoids extraneous
conversion nodes.
A new function promoteAggregate is added, and used by findFunction. This is essentially
a generalization of the "promote 1st or 2nd arg" algorithm in promoteBinary.
The actual selection proceeds in three steps, as described in the comments in
hlslParseContext::findFunction:
1. Attempt an exact match. If found, use it.
2. If not, obtain the operator from step 1, and promote arguments.
3. Re-select the intrinsic overload from the results of step 2.
HLSL has keywords for various interpolation modifiers such as "linear",
"centroid", "sample", etc. Of these, "sample" appears to be special,
as it is also accepted as an identifier string, where the others are not.
This PR adds this ability, so the construct "int sample = 42;" no longer
produces a compilation error.
New test = hlsl.identifier.sample.frag
