Makes codegen from typical D3D emulation SPIR-V more readable.
Also makes cross compilation with NotEqual more sensible.
It's very rare to actually need the strict NaN-checks in practice.
Also, glslang now emits UnordNotEqual by default it seems, so give up
trying to assume OrdNotEqual. Harmonize for UnordNotEqual as the sane
default.
SPIR-V allows an image to be marked as a depth image, but with a non-depth
format. Such images should be read or sampled as vectors instead of scalars,
except when they are subject to compare operations.
Don't mark an OpSampledImage as using a compare operation just because the
image contains a depth marker. Instead, require that a compare operation
is actually used on that image.
Compiler::image_is_comparison() was really testing whether an image is a
depth image, since it incorporates the depth marker. Rename that function
to is_depth_image(), to clarify what it is really testing.
In Compiler::is_depth_image(), do not treat an image as a depth image
if it has been explicitly marked with a color format, unless the image
is subject to compare operations.
In CompilerMSL::to_function_name(), test for compare operations
specifically, rather than assuming them from the depth-image marker.
CompilerGLSL and CompilerMSL still contain a number of internal tests that
use is_depth_image() both for testing for a depth image, and for testing
whether compare operations are being used. I've left these as they are
for now, but these should be cleaned up at some point.
Add unit tests for fetch/sample depth images with color formats and no compare ops.
New in MSL 2.3 is a template that can be used in the place of a scalar
type in a stage-in struct. This template has methods which interpolate
the varying at the given points. Curiously, you can't set interpolation
attributes on such a varying; perspective-correctness is encoded in the
type, while interpolation must be done using one of the methods. This
makes using this somewhat awkward from SPIRV-Cross, requiring us to jump
through a bunch of hoops to make this all work.
Using varyings from functions in particular is a pain point, requiring
us to pass the stage-in struct itself around. An alternative is to pass
references to the interpolants; except this will fall over badly with
composite types, which naturally must be flattened. As with
tessellation, dynamic indexing isn't supported with pull-model
interpolation. This is because of the need to reference the original
struct member in order to call one of the pull-model interpolation
methods on it. Also, this is done at the variable level; this means that
if one varying in a struct is used with the pull-model functions, then
the entire struct is emitted as pull-model interpolants.
For some reason, this was not documented in the MSL spec, though there
is a property on `MTLDevice`, `supportsPullModelInterpolation`,
indicating support for this, which *is* documented. This does not appear
to be implemented yet for AMD: it returns `NO` from
`supportsPullModelInterpolation`, and pipelines with shaders using the
templates fail to compile. It *is* implemeted for Intel. It's probably
also implemented for Apple GPUs: on Apple Silicon, OpenGL calls down to
Metal, and it wouldn't be possible to use the interpolation functions
without this implemented in Metal.
Based on my testing, where SPIR-V and GLSL have the offset relative to
the pixel center, in Metal it appears to be relative to the pixel's
upper-left corner, as in HLSL. Therefore, I've added an offset 0.4375,
i.e. one half minus one sixteenth, to all arguments to
`interpolate_at_offset()`.
This also fixes a long-standing bug: if a pull-model interpolation
function is used on a varying, make sure that varying is declared. We
were already doing this only for the AMD pull-model function,
`interpolateAtVertexAMD()`; for reasons which are completely beyond me,
we weren't doing this for the base interpolation functions. I also note
that there are no tests for the interpolation functions for GLSL or
HLSL.
DXVK emits SPIR-V where fragment shader builtins have names derived from
DXBC assembly, e.g. `oDepth` for `FragDepth`. When we declared the
disabled output, we used this name, but when referencing it, we
continued to use the GLSL name. This breaks compilation.
Fix fallout from changes.
There's a bug in glslang that prevents `float16_t`, `[u]int16_t`, and
`[u]int8_t` constants from adding the corresponding SPIR-V capabilities.
SPIRV-Tools, meanwhile, tightened validation so that these constants are
only valid if the corresponding `Float16`, `Int16`, and `Int8` caps are
on. This affects the `16bit-constants.frag` test for GLSL and MSL.
Just like OpAccessChain we need to make use of the meta information
available to use from access_chain_internal as we can extract a packed
vector or transposed vector from a composite, not just memory load.
- Add new Windows support
- Use CMake/CTest instead of Make + shell scripts
- Use --parallel in CTest
- Fix CTest on Windows
- Cleanups in test_shaders.py
- Force specific commit for SPIRV-Headers
- Fix Inf/NaN odd-ball case by moving to ASM
A lot of changes in spirv-opt output.
Some new invalid SPIR-V was found but most of them were not significant
for SPIRV-Cross, so just marked them as invalid.
This roughly matches their semantics in SPIR-V and MSL. For `FMin`,
`FMax`, and `FClamp`, and the Metal functions `fast::min()`,
`fast::max()`, and `fast::clamp()`, the result is undefined if any
operand is NaN. For the 'N' operations and their corresponding MSL
`precise::` functions, the result is consistent with IEEE 754 (first
non-NaN wins; result is NaN if all operands are NaN).
We can only do this with 32-bit floats, though, because Metal only
provides these variants for `float`. `half` only has one variant of
these functions that is presumably consistent with IEEE 754. I guess
that's OK; the SPIR-V spec only says that `F{Min,Max,Clamp}` are
undefined for NaNs. Performance might suffer, though.
The SPIR-V spec says that these check if the operands either are
unordered or satisfy the given condition. So that's just what we'll do,
using Metal's `isunordered()` stdlib function. Apple's optimizers ought
to be able to collapse that to a single unordered compare.
MSL would force thread const& which would not work if the input argument
came from a different storage class.
Emit proper non-reference arguments for such values.
Certain patterns with OpVectorShuffle (and probably others) will cascade
to so large, that they can cause OOM. After we have observed
force_recompile, don't spend unnecessary memory emitting code which will
never be used.
Normally, temporary declaration must dominate any use of it,
so we generally did not need to analyze the CFG for these variables,
but there is an edge case where you have an inliner doing:
do {
create_temporary;
break;
} while(0);
use_temporary;
The inside of the loop dominates the outer scope, but we cannot emit
code like this in GLSL, so make sure we hoist these temporaries outside
the "loop".
