A struct declaration with variables is now absorbed into the 'declaration'
rule, like any other variable declaration.
A struct declaration without variables is now reduced to the
'struct_declaration_without_vars' rule.
They both are reduced to a 'declaration_statement' in the end.
In a declaration with multiple variables, the variables must be created
before the initializer of the next variable is parsed. This is required
for initializers such as:
float a = 1, b = a, c = b + 1;
A requisite for this is that the type information is parsed in the same
rule as the first variable (as a variable_def_typed) so it is
immediately available to declare the first variable. Then, the next
untyped variable declaration is parsed, and the type from the first
variable can be used to declare the second, before the third is parsed,
and so on.
Non-constant vector indexing is not solved with relative addressing
in the register indexes because this indexation cannot be at the level
of register-components.
Mathematical operations must be used instead.
Currently, the compiler requires that dereferences be HLSL_IR_CONSTANT, so that
it can compute the offset at compile time. However, scenarios such as this test
will produce a dereference with HLSL_IR_EXPR, which will generate an error.
Passing this test in particular will require adding support for SM4 relative
addressing, as well as support for non-constant indexing in general.
Signed-off-by: Ethan Lee <flibitijibibo@gmail.com>
In Shader Model 6 each signature element can span a range of register
indices, or 'rows', and system values do not share a register index with
non-system values. Inputs and outputs are referenced by element index
instead of register index. This patch merges multiple signature elements
into a single element under the following conditions:
- The register index in a load or store is specified dynamically by
including a relative address parameter with a base register index. The
dcl_index_range instruction is used to identify these.
- A register declaration is split across multiple elements which declare
different components of the register.
- A patch constant function writes tessellation factors. These are an
array in SPIR-V, but in SM 5.x each factor is declared as a separate
register, and these are dynamically indexed by the fork/join instance
id. Elimination of multiple fork/join phases converts the indices to
constants, but merging the signature elements into a single arrayed
element matches the SPIR-V output.
All references to input/output register indices are converted to element
indices. If a relative address is present, the element index is moved up
a slot so it cannot be confused with a constant offset. Existing code
only handles register index relative addressing for tessellation factors.
This patch adds generic support for it.
The new fixmes can be triggered in presence of object components within
structs (for SM5).
In shaders such as this one:
struct apple
{
Texture2D tex : TEX;
float4 color : COLOR;
};
float4 main(struct apple input) : sv_target
{
return input.tex.Load(int3(1, 2, 3));
}
Or this one:
struct
{
Texture2D tex;
float4 color;
} s;
float4 main() : sv_target
{
return s.tex.Load(int3(1, 2, 3));
}
Variables that contain more than one object (arrays or structs) require
the allocation of contiguous registers in the respective object
register spaces.
Otherwise, in the added test, we get:
vkd3d-compiler: vkd3d-shader/hlsl.c:452: hlsl_init_deref_from_index_chain: Assertion `chain' failed.
because on the path that triggers the following error:
E5002: Wrong type for argument 1 of 'tex3D': expected 'sampler' or 'sampler3D', but got 'sampler2D'.
a NULL params.resource is passed to hlsl_new_resource_load() and
then to hlsl_init_deref_from_index_chain().
The use of the hlsl_semantic.reported_duplicated_output_next_index field
allows reporting multiple overlapping indexes, such as in the following
vertex shader:
void main(out float1x3 x : OVERLAP0, out float1x3 y : OVERLAP1)
{
x = float3(1.0, 2.0, 3.2);
y = float3(5.0, 6.0, 5.0);
}
apple.hlsl:1:41: E5013: Output semantic "OVERLAP1" is used multiple times.
apple.hlsl:1:13: First use of "OVERLAP1" is here.
apple.hlsl:1:41: E5013: Output semantic "OVERLAP2" is used multiple times.
apple.hlsl:1:13: First use of "OVERLAP2" is here.
While at the same time avoiding reporting overlaps more than once for
large arrays:
struct apple
{
float2 p : sv_position;
};
void main(out apple aps[4])
{
}
apple.hlsl:3:8: E5013: Output semantic "sv_position0" is used multiple times.
apple.hlsl:3:8: First use of "sv_position0" is here.
Thanks to Giovanni for the second set of tests! Note that the
tolerance for the final pixel was set much higher than the others;
this test seems to be an issue for some devices (in my case, a 7900
XTX running RADV).
Co-authored-by: Giovanni Mascellani <gmascellani@codeweavers.com>
Signed-off-by: Ethan Lee <flibitijibibo@gmail.com>
From this point on, it is no longer true that only hlsl_ir_loads can
return objects, because an object can also come from chain of
hlsl_ir_indexes that ends in an hlsl_ir_load.
The lower_index_loads pass takes care of lowering all hlsl_ir_indexes
into hlsl_ir_loads.
For this reason, hlsl_resource_load_params now expects both the resource
as the sampler to be just an hlsl_ir_node pointer instead of a pointer
to a more specific hlsl_ir_load.
Some drivers (AMD Radeon RX 6700 XT, with radeonsi from Mesa 22.2.0-rc3) emit
less than one invocation per pixel, presumably because they detect that the
shader control flow is uniform for all pixels. Having the control flow depend on
SV_Position avoids this test failure.
Cf. 34bd0dd0704c613abef8a9aa3ba2a2507ed02843 in wine.
The expected use case where a heap is freed before its contained
resources is not reasonably testable, so the ability to place a new
resource is tested instead.
But still throw hlsl_fixme() when there is more than one.
Prioritizing among multiple compatible function overloads in the same way
as the native compiler would require systematic testing.
This was originally left alone in order to allow functions without early return
to succeed, since in that case we would already emit the correct bytecode
despite not handling the HLSL_IR_JUMP_RETURN instruction.
Now that we lower return statements, however, any unhandled instructions are
either definitely going to result in invalid bytecode, or rare enough that it's
not worth returning success anyway.
Vectors cannot be used as array indexes, however, single-component
swizzles (such as vec.x) can be used.
This suggests that single-component swizzles should actually be
scalars and not vectors of dimx = 1.
It is worth noting that the use of single-component swizzles on scalars
should still be allowed.
Co-authored-by: Francisco Casas <fcasas@codeweavers.com>
Co-authored-by: Zebediah Figura <zfigura@codeweavers.com>
Because copy_propagation_transform_object_load() replaces a deref
instead of an instruction, it is currently prone to two problems:
1- It can replace a deref with the same deref, returning true every
time and getting the compilation stuck in an endless loop of
copy-propagation iterations.
2- When performed multiple times in the same deref, the second time it
can replace the deref with a deref from a temp that is only valid in
another point of the program execution, resulting in an incorrect value.
This patch preempts this by avoiding replacing derefs when the new deref
doesn't point to a uniform variable. Because, uniform variables cannot
be written to.
Reinterpret min16float, min10float, min16int, min12int, and min16uint
as their regular counterparts: float, float, int, int, uint,
respectively.
A proper implementation would require adding minimum precision
indicators to all the dxbc-tpf instructions that use these types.
Consider the output of fxc 10.1 with the following shader:
uniform int i;
float4 main() : sv_target
{
min16float4 a = {0, 1, 2, i};
min16int2 b = {4, i};
min10float3 c = {6.4, 7, i};
min12int d = 9.4;
min16uint4x2 e = {14.4, 15, 16, 17, 18, 19, 20, i};
return mul(e, b) + a + c.xyzx + d;
}
However, if the graphics driver doesn't have minimum precision support,
it ignores the minimum precision indicators and runs at 32-bit
precision, which is equivalent as working with regular types.
If a hlsl_ir_load loads a variable whose components are stored from different
instructions, copy propagation doesn't replace it.
But if all these instructions are constants (which currently is the case
for value constructors), the load could be replaced with a constant value.
Which is expected in some other instructions, e.g. texel_offsets when
using aoffimmi modifiers.
For instance, this shader:
```
sampler s;
Texture2D t;
float4 main() : sv_target
{
return t.Gather(s, float2(0.6, 0.6), int2(0, 0));
}
```
results in the following IR before applying the patch:
```
float | 6.00000024e-01
float | 6.00000024e-01
uint | 0
| = (<constructor-2>[@4].x @2)
uint | 1
| = (<constructor-2>[@6].x @3)
float2 | <constructor-2>
int | 0
int | 0
uint | 0
| = (<constructor-5>[@11].x @9)
uint | 1
| = (<constructor-5>[@13].x @10)
int2 | <constructor-5>
float4 | gather_red(resource = t, sampler = s, coords = @8, offset = @15)
| return
| = (<output-sv_target0> @16)
```
and this IR afterwards:
```
float2 | {6.00000024e-01 6.00000024e-01 }
int2 | {0 0 }
float4 | gather_red(resource = t, sampler = s, coords = @2, offset = @3)
| return
| = (<output-sv_target0> @4)
```
On cross builds, shaders are compiled with d3dcompiler_47.dll and
run with d3dN.dll. On non-cross builds, shaders are compiled with
vkd3d-shader and run with d3dN.dll (on Windows) or Vulkan and vkd3d
(on Linux).
validate_static_object_references() validates that uninitialized static
objects are not referenced in the shader.
In case a static variable contains both numeric and object types, the
"Static variables cannot have both numeric and resource components."
error should preempt uninitialized numeric values to reach further
compilation steps.
Note that in the future we should call
validate_static_object_references() after DCE and pruning branches,
because shaders such as these compile (at least in more modern versions
of the native compiler):
Branch pruning:
```
static RWTexture2D<float> tex;
float4 main() : sv_target
{
if (0)
{
tex[int2(0, 0)] = 2;
}
return 0;
}
```
DCE:
```
static Texture2D tex;
uniform uint i;
float4 main() : sv_target
{
float4 unused = tex.Load(int3(0, 1, 2));
return 0;
}
```
These are "todo" tests in hlsl-static-initializer.shader_test
that depend on this.
We are currently not initializing static values to zero by default.
Consider the following shader:
```hlsl
static float4 va;
float4 main() : sv_target
{
return va;
}
```
we get the following output:
```
ps_5_0
dcl_output o0.xyzw
dcl_temps 2
mov r0.xyzw, r1.xyzw
mov o0.xyzw, r0.xyzw
ret
```
where r1.xyzw is not initialized.
This patch solves this by assigning the static variable the value of an
uint 0, and thus, relying on complex broadcasts.
This seems to be the behaviour of the 9.29.952.3111 version of the native
compiler, since it retrieves the following error on a shader that lacks
an initializer on a data type with object components:
```
error X3017: cannot convert from 'uint' to 'struct <unnamed>'
```
Using add_unary_arithmetic_expr() instead of hlsl_new_unary_expr()
allows the intrinsic to work with matrices.
Otherwise we get:
E5017: Aborting due to not yet implemented feature: Copying from unsupported node type.
because an HLSL_IR_EXPR reaches split_matrix_copies().
Some intrinsics have different rules for the allowed data types than
expressions:
- Vectors and matrices at the same time are not allowed, regardless of
their dimensions. Even if they have the same number of components.
- Any combination of matrices is always allowed, even those when no
matrix fits inside another, e.g.:
float2x3 is compatible with float3x2, resulting in float 2x2.
The common data type is the min on each dimension.
This is the case for max, pow, ldexp, clamp and smoothstep; which suggest that
it is the case for all intrinsics where the operation is applied element-wise.
Tests for mul() are also added as a counter-example where the operation
is not element-wise.
Until vkd3d-shader is patched, an atomic op on a typed buffer where
StorageImageReadWithoutFormat is available will cause SPIR-V validation
failure, and assertion in Mesa debug builds, because the image will be
declared with Unknown format.
Otherwise, for instance, the added test results in:
debug_hlsl_writemask: Assertion `!(writemask & ~VKD3DSP_WRITEMASK_ALL)' failed.
Which happens in allocate_variable_temp_register() when the variable's
type reg_size is <= 4 but its component count is larger, which may
happen if it contains objects.
Do not rely on a draw or dispatch command to do this.
This allows more efficiently testing syntax, in cases where testing the actual
shader functionality is not interesting.
Otherwise the test fails on a NVIDIA GeForce GTX 1050 Ti GPU.
The error being:
shader_runner:535:Section [test], line 9: Test failed: Got {2.72165507e-01, 4.08248246e-01, 5.44331014e-01, 6.80413783e-01}, expected {2.72165537e-01, 4.08248305e-01, 5.44331074e-01, 6.80413842e-01} at (0, 0).
This should silence warnings about some branches non returning any value
without requiring additional "return 0" statement or similar.
Also, in theory this might enable to compiler to optimize the program
a little bit more, though that's unlikely to have any measurable effect.
HLSL_ARRAY_ELEMENTS_COUNT_IMPLICIT (zero) is used as a temporal value
for elements_count for implicit size arrays.
This value is replaced by the correct one after parsing the initializer.
In case the implicit array is not initialized correctly, hlsl_error()
is called but the array size is kept at 0. So the rest of the code
must handle these cases.
In shader model 5.1, unlike in 5.0, declaring a multi-dimensional
object-type array with the last dimension implicit results in
an error. This happens even in presence of an initializer.
So, both gen_struct_fields() and declare_vars() first check if the
shader model is 5.1, the array elements are objects, and if there is
at least one implicit array size to handle the whole type as an
unbounded resource array.
Signed-off-by: Francisco Casas <fcasas@codeweavers.com>
It is responsibility of the shader's programmer to ensure that
object references can be solved statically.
Resource arrays for ps_5_1 and vs_5_1 are an exception which is not
properly handled yet. They probably deserve a different object type.
Signed-off-by: Francisco Casas <fcasas@codeweavers.com>
Otherwise we get false in implicit_compatible_data_types() when passing
types that are equal but not convertible according to
convertible_data_type(); e.g. getting:
"Can't implicitly convert from Texture2D<float4> to Texture2D<float4>."
Signed-off-by: Francisco Casas <fcasas@codeweavers.com>
Fixes reflections in Control appearing with only their red component.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52146
Signed-off-by: Conor McCarthy <cmccarthy@codeweavers.com>
This currently fails if the shader loads from the UAV, because it causes
vkd3d-shader to specify the R32f format instead of Unknown.
Signed-off-by: Conor McCarthy <cmccarthy@codeweavers.com>
Prepare to allow for dynamically changing the bound attachments in consecutive
draw calls.
Signed-off-by: Giovanni Mascellani <gmascellani@codeweavers.com>
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Split the probe_vec4() directive into get_rt_readback() and release_readback().
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Giovanni Mascellani <gmascellani@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Move the "resource" field to a new "d3d12_resource_readback" structure
encapsulating struct resource_readback.
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Giovanni Mascellani <gmascellani@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Signed-off-by: Francisco Casas <fcasas@codeweavers.com>
Signed-off-by: Giovanni Mascellani <gmascellani@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Zebediah Figura <zfigura@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Strictly increasing timeline values must be mapped to fence virtual values
to avoid invalid use of Vulkan timeline semaphores. In particular, non-
increasing values and value jumps of >= 4G are permitted in d3d12.
Different virtual D3D12 command queues may map to the same Vulkan queue.
If a wait of value N is submitted on one command queue, and then a signal
for >= N is submitted on another, but they are sent to the same Vk queue,
the wait will never complete. The solution is to buffer out-of-order waits
and any subsequent queue commands until an unblocking signal value is
submitted to a different D3D12 queue, or signaled on the CPU.
Buffering out-of-order waits also fixes the old fence implementation so it
is fully functional, though a bit less efficient than timeline semaphores.
Based in part on vkd3d-proton patches by Hans-Kristian Arntzen. Unlike the
vkd3d-proton implementation, this patch does not use worker threads for
submissions to the Vulkan queue.
Signed-off-by: Conor McCarthy <cmccarthy@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
