Fixes#1305.
Ensure that two `back::spv::LocalType::Image` values are sure to be equal (and
hash equal) whenever they would generate the same `OpTypeImage` instruction, so
that the usual duplicate removal via `Writer::lookup_type` works. Accomplish
this by changing the contents of `LocalType::Image` to more closely match the
operands of `OpTypeImage` instructions.
Previously, `LocalType::Image` included a `ImageClass::Storage::access` value,
which did not affect the `OpTypeImage` instruction generated. If two
`LocalType::Image` values differered only in their `access`, then they would get
separate entries in `Writer::lookup_type`, two identical `OpTypeImage`
instructions would be generated, and SPIR-V validation would fail.
Change `write_type_declaration_local` to handle any type that can be represented
as a `LocalType`. This means that Image and Sampler types are handled there now.
Change `write_type_declaration_arena` to defer to `write_type_declaration_local`
to handle any type that can be represented as a `LocalType`. This makes the code
for those types in `write_type_declaration_arena` redundant, so remove it.
However, continue to request SPIR-V capabilities for image types from
`write_type_declaration_arena`, since that function is fallible, while
`write_type_declaration_local` is not.
When we are leaning on robust buffer access to do the job for us, there's no
undefined behavior going on. So `UndefinedBehavior` suggests people are doing
something reckless even if they're not. The policy just says what Naga is doing,
and it shouldn't pretend to say what the rest of the system is doing.
* Implement lexing for all WGSL number literal types
* Move number literal test cases
* Adjust tests to match WGSL spec on number literals
Suffixes are not type names and currently only a plain `u` is supported
for uints. More specifically, `i` and `f` suffixes or suffixes with
widths in bits like `u32` are not supported at the moment.
* Add more tests for invalid number literal suffixes
* Replace code too new for Rust 1.43
* Implement parsing for hexadecimal integers
* Switch to enum number types, and Bytes for width
* Check for negative and leading zeros in int literals
* Implement parsing of hex floats with hexf-parse
* Update error message tests
* Update snapshot test output files
* Clean up lexer state machine code
* Clean up unexpected token error code
* Move number literal parsing to own submodule
Using 1D images should require either the `Sampled1D` or `Image1D` capability.
Using cube array images should require either the `SampledCubeArray` or
`ImageCubeArray` capability.
Introduce a helper type that manages the details of emitting SPIR-V
conditionals, tracking labels, branches, merge annotations and phi nodes.
Change index bounds checks to use this helper.
For ease of review, there should be no meaningful code changes in this commit.
Match arms are merely moved out into their own functions, whose arguments are
the enum variant's fields, along with any free variables from the function
containing the match.
* [hlsl-out] flatten the entry point inputs
Previously, the logic was re-ordering the inputs according to the binding.
This breaks if one of the inputs is a struct. With this change, the struct
fields are also flattened into the fake entry point struct. We also
construct the original arguments at the beginning of the function.
* hlsl-out: completely separate the flattened IO structs from the original IR structs
Previously, we had heuristics to detect if a particular struct needs the fields
to be re-ordered. We'd re-order interface structs without layout, and the detection
was very fragile and easily wrong.
The new logic is spawning separate struct types if we need any re-ordering to happen.
It's solid, there are no heuristics.
Remove `forbidden_caps`.
Accumulate capabilities actually used separately from the permitted
capabilities, so that the latter can be retained across Writer resets, while the
former is cleared between modules.
Hlsl and wgsl don't support them directly so a polyfill is used taken
from the msl spec.
`asinh` = `log(x + sqrt(x * x + 1.0))`
`acosh` = `log(x + sqrt(x * x - 1.0))`
`atanh` = `0.5 * log((1.0 + x) / (1.0 – x))`
