Replace uses of `call_unique` with uses of `call` and `call_or`, which becomes
public. It's not clear when `call_unique` is correct to use, and avoiding a few
numeric suffixes here and there isn't worth it.
Eliminate `Namer::namespace_index` and remove namespace indices from
`Namer::unique` keys. Instead, implement `Namer::namespace` by just swapping in
a fresh table for the duration of the call.
Create the fresh hash table with the right initial capacity. The prior
implementation did manage to avoid allocations by sharing a hash table, and we'd
like to not lose that advantage entirely.
This has no effect on generated code.
Treat `TypeInner::ValuePointer` and `TypeInner::Pointer` as equivalent by
converting them to a canonical form before comparison.
Support `ValuePointer` in WGSL type output.
Fixes#1318.
Ensure that each distinct type occurs only once in `Module::types`, so that we
can use `Eq` on `Type` or `TypeInner` for type equivalence, without being
confused by differing `Handle<Type>` values that point to identical types.
This removes a number of duplicate types from the ir snapshots.
Fixes#1385.
Replace `Module::apply_common_default_interpolation` with a simpler function
that handles a single `Binding` at a time. In exchange for the simplicity, the
function must be called at each point function arguments, function results, and
struct members are prepared. (Any missed spots will be caught by the verifier.)
This approach no longer requires mutating types in the arena, a prerequisite for
properly handling type identity.
Applying defaults to struct members when the struct declaration is parsed does
have a disadvantage, compared to the old whole-module pass: at struct parse
time, we don't yet know which pipeline stages the struct will be used in. The
best we can do is apply defaults to anything with a `Location` binding. This
causes needless qualifiers to appear in some output. However, it seems that our
back end languages all tolerate such qualifiers.
Previously the typifier flagged all relational functions as producing a
scalar boolean but with the exception of `all` and `any`, according to
the wgsl spec, all other relational functions output a type the same
size as the input.
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))`
* Proof of concept for "span" feature, with WGSL parsing augmented.
* Review:
1) add_span was actually a bad idea, make it set_span and add
set_span_if_unknown too.
2) panics on getting/setting span for invalid handles.
3) only set span for constants with a name
4) don't overwrite spans for types.
* Added spans to blocks & more expressions getting spans in frontends.
Definitely the shotgunny type of commit, but what can you do. The design
I went with made spans mandatory to specify, so I had to go and wire
them through wherever I could.
* Moved Block to a separate module, +clippy
* More spans for types in GLSL.
* Remove pointless body method.
* Make Arena interface require spans.
Another shotgun commit, oh boy...
* Fix tests.
My loathsome habit to "quickly fix things along the way" made a lot of
extra work for me here, having to fix my "fixes" for WGSL parser.
* Rustfmt + clippy.
* Fix compile-errors with span feature enabled.
* Nuked set_span* from orbit. Deleting code feels great!
* Code review - move feature flags inside functions.
* Fix build with "deserialize" feature enabled.
* Resurrect texture_storage_* tests
* Test parsing of `var<storage,write>`
* Default storage textures to READ
* Restore default features
* Fix glsl/hlsl/msl/spv front and back ends
* Add missing test outputs
* All-around fixes for the storage access
Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
This makes Naga IR validation impose the restrictions added to WGSL in
gpuweb/gpuweb#1801.
Remove code in the SPIR-V writer to spill arrays to temporary variables in order
to index them dynamically. If such IR is encountered, treat it as a failure of
validation.
The `apply_common_default_interpolation` helper function would panic if bindings
were missing, but missing bindings should be something that front ends can count
on validation to detect, so the helper should just return silently.
The validator returned `InvalidType` errors for missing bindings, apparently
because variables without bindings must be structs that do have bindings. But
this is unhelpful when you've just forgotten to label an argument. So this patch
adds a new, more specific, `VaryingError` variant.
When validating IR, verify that all `Binding`s for vertex shader outputs and
fragment shader inputs, whether directly in the argument or result, or applied
to a struct member, has specified an interpolation and sampling, not `None`.
This ensures that front ends explicitly state their policies, rather than
coasting through on back ends' default behavior.
In practice, all our front ends have very similar defaults, so provide a utility
function on `Module` to apply these rules. Use this utility function in the
SPIR-V and WGSL front ends; GLSL seems to already fill in interpolation as
required.
