Commit Graph
412 Commits
Author SHA1 Message Date
Connor Fitzgerald 57d787eb53 Slang Binding Array Test 2024-10-20 23:50:47 -04:00
dependabot[bot]andConnor Fitzgerald e06f10e0a4 build(deps): bump crate-ci/typos from 1.24.6 to 1.26.0 (#6399)
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
2024-10-21 00:23:43 +00:00
Erich Gubler b3f665be7d feat(wgsl-in): create skeleton for parsing directives 2024-10-18 16:08:46 -04:00
Erich Gubler cb31465811 refactor(wgsl-in): add Lexer::peek_ident_with_span
1. Break out `word_as_ident*` helpers to keep validation of identifiers
   DRY.
2. Add `peek_*` variant of `Lexer::next_ident_with_span`.

This will be consumed immediately in the subsequent commit.
2024-10-18 16:08:46 -04:00
Jasper St. Pierre 1b2ef8612d naga: Add support for textureQueryLevels to GLSL parser (#6415) 2024-10-18 10:07:58 +02:00
Xiaopeng Li 74ef445bca ADD metal namespace for uint4 (#6417) 2024-10-17 11:45:29 +02:00
Erich Gubler f2faf609af refactor(glsl-out): hoist new if let TypeInner::Struct out of varying_required_features (#6230) 2024-10-14 12:45:32 +02:00
FL33TW00DandErich Gubler f842d80f52 chore: note: round_ties_even is on stable, but not MSRV yet
Co-Authored-By: Erich Gubler <erichdongubler@gmail.com>
2024-10-14 06:43:53 -04:00
Erich Gubler a13e710956 chore: satisfy unused_qualifications (again)
Regressed by
[#6354](https://github.com/gfx-rs/wgpu/pull/6354/files#diff-1dd0973ceb242a02442d33e422f9dff40e007b505d0ee726613e1763e61e768dR2809)
(CC @jimb). Hopefully will get plugged by
[#6105](https://github.com/gfx-rs/wgpu/pull/6105/).
2024-10-11 12:27:34 -07:00
Erich Gubler d22aa2eb55 chore: format TOML files with taplo 0.9.3 2024-10-11 14:40:01 -04:00
Schell Carl Scivally 76b1605090 [spv-in] sampling from depth textures returns a scalar and is splatted (#6384) 2024-10-11 20:24:16 +02:00
Jim Blandy 1047fa57f0 [naga] Move array-by-value snapshot tests to index-by-value.wgsl. 2024-10-11 08:27:15 -07:00
Jim Blandy ed3006ccc6 [naga spv-out] Spill arrays and matrices for runtime indexing.
Improve handling of `Access` expressions whose base is an array or
matrix (not a pointer to such), and whose index is not known at
compile time. SPIR-V does not have instructions that can do this
directly, so spill such values to temporary variables, and perform the
accesses using `OpAccessChain` instructions applied to the
temporaries.

When performing chains of accesses like `a[i].x[j]`, do not reify
intermediate values; generate a single `OpAccessChain` for the entire
thing.

Remove special cases for arrays; the same code now handles arrays and
matrices.

Update validation to permit dynamic indexing of matrices.

For details, see the comments on the new tracking structures in
`naga::back::spv::Function`.

Add snapshot test `index-by-value.wgsl`.

Fixes #6358.
Fixes #4337.
Alternative to #6362.
2024-10-11 08:27:15 -07:00
Jim Blandy 475a716822 [naga spv-out] Let write_checked_load take AccessTypeAdjustment.
Let `BlockContext::write_checked_load` take an `AccessTypeAdjustment`
argument, so that the caller can choose what adjustment to apply to
`pointer`.
2024-10-11 08:27:15 -07:00
Jim Blandy 26a95fd270 [naga spv-out] Add some tracing output to Writer::write_function. 2024-10-11 08:27:15 -07:00
Jim Blandy 436ffba77a [naga spv-out] Introduce Writer::get_resolution_pointer_id.
Introduce a new helper function,
`naga::back::spv::Writer::get_resolution_pointer_id`. Use it in
`BlockContext::write_expression_pointer`.
2024-10-11 08:27:15 -07:00
Jim Blandy 0f17ad6455 [naga] Add new function, GuardedIndex::from_expression.
Pull out the code to build a `naga::proc::index::GuardedIndex` from a
`Handle<Expression>` into its own function,
`GuardedIndex::from_expression`. Use that function in
`GuardedIndex::try_resolve_to_constant`.
2024-10-11 08:27:15 -07:00
Jim Blandy 0392613b5a [naga spv-out] Abstract out NumericType::from_inner.
Pull out the code for building a `naga::back::spv::NumericType` from a
`TypeInner` into its own function, `NumericType::from_inner`. Use that
in `LocalType::from_inner`.
2024-10-11 08:27:15 -07:00
Jim Blandy 4427ff9622 [naga spv-out] Use crate::proc::index::GuardedIndex.
Replace qualified paths with a `use` directive.
2024-10-11 08:27:15 -07:00
Jim Blandy 57b8858f96 [naga spv-out] Gather array, matrix, and vector cases.
This commit is just code motion.
2024-10-11 08:27:15 -07:00
Jim Blandy f9075fc4b8 [naga] Test access to a member/element through a pointer. 2024-10-11 08:27:15 -07:00
Jim Blandy b9f1e4a266 [naga spv-out] Clean up write_expression_pointer type adjustment.
Replace the `return_type_override` argument of
`BlockContext::write_expression_pointer` with an enum that says how to
derive the return type from `expr_handle`'s type.

Introduce a new type, `AccessTypeAdjustment`, that covers possible
derivation rules.

This simplifies callers and the callee, in part by making the possible
alternatives less general, and by giving them explicit names (the
variants of the `AccessTypeAdjustment` enum).
2024-10-11 08:27:15 -07:00
Jim Blandy d034c4b428 [naga spv-out] Move code to load a pointer into its own function.
Introduce a new function,
`naga::back::spv::BlockContext::write_checked_load`, that does the
work of `Expression::Load`.

This change is just code motion, and should have no effect on
behavior. The new function will be used in later commits.
2024-10-11 08:27:15 -07:00
Ronny Chan 73764fdc6a [naga/wgsl-out]: polyfill inverse function (#6385) 2024-10-11 15:56:12 +02:00
Jim Blandy ae52e5dc96 [naga spv-out] Delete BlockContext::is_intermediate; use types.
Delete the function `BlockContext::is_intermediate`. Instead, have
`Access` and `AccessIndex` instructions decide whether to defer code
generation based on the type of the base expression: indexing
operations on pointers are deferred; anything else is not.
2024-10-10 07:42:02 -07:00
Jim Blandy 5c6b00886e [naga spv-out] Simplify Writer::get_pointer_id.
Simplify the definition of `naga::back::spv::Writer::get_pointer_id`
by using `get_type_id`'s ability to handle `LocalType::Pointer`.

This means that `get_pointer_id` is no longer fallible, and no longer
needs a type arena. Simplify callers, as well as the
`BlockContext::get_pointer_id` convenience function.
2024-10-10 07:41:35 -07:00
Jim Blandy 908e8353a8 [naga spv-out] Expand LocalType to permit pointers to matrices.
In `back::spv`:

- Factor out the numeric variants of `LocalType` into a
  new enum, `NumericType`.

- Split the `Value` variant into `Numeric` and `LocalPointer`
  variants, and let `LocalPointer` point to any numeric type,
  including matrices.

In subsequent commits, we'll need to spill matrices out into temporary
local variables. This means we'll need to generate SPIR-V
pointer-to-matrix types, so `LocalType` needs to be able to represent
that.
2024-10-10 07:41:15 -07:00
Jim Blandy 0392cb783d [naga spv-out] Rename make_local to LocalType::from_inner.
Change the free function `back::spv::make_local` into an associated
function `LocalType::from_inner`.
2024-10-10 07:41:15 -07:00
Asher Jingkong Chen bf33e481f3 [naga msl-out] Implement atomicCompareExchangeWeak for MSL backend (#6265) 2024-10-10 12:45:24 +02:00
Jim Blandy 3693da27f9 [naga spv-out] Bounds-check runtime-sized array access correctly.
Do not neglect to apply bounds checks to indexing operations on
runtime-sized arrays, even when they are accessed via an `AccessIndex`
instruction.

Before this commit, `BlockContext::write_expression_pointer` would not
apply bounds checks to `OpAccessChain` indices provided by an
`AccessIndex` instruction, apparently with the rationale that any
out-of-bounds accesses should have been reported by constant
evaluation.

While it is true that the `index` operand of an `AccessIndex`
expression is known at compile time, and that the WGSL constant
evaluation rules require accesses that can be statically determined to
be out-of-bounds to be shader creation or pipeline creation time
errors, accesses to runtime-sized arrays don't follow this pattern:
even if the index is known, the length with which it must be compared
is not.

Fixes #4441.
2024-10-08 11:53:15 -07:00
Jim Blandy 7283185305 [naga] Extend snapshot tests for bounds checks.
Extend the snapshot tests for bounds checking to cover the case where
a runtime-sized array is indexed by a constant.
2024-10-08 11:53:15 -07:00
Jim Blandy 5a6b749335 [naga spv-out] Let BoundsCheckResult supply the index value.
Let `BoundsCheckResult::Conditional` provide both the condition to
check before carrying out the access, and the index to use for that
access. The `Conditional` variant indicates that we generated a
runtime bounds check, which implies we must have had a SPIR-V id for
the index to pass to that check, so there's no reason not to provide
that to the callers - especially if the bounds check code was able to
reduce it to a known constant.

At the moment, this is not much of a refactor, but later commits will
use `GuardedIndex` in more places, at which point this will avoid a
re-matching and assertion.
2024-10-08 11:53:15 -07:00
Jim Blandy 69ab63ca34 [naga spv-out] Abstract out OpAccessChain index production.
Abstract the code from `write_expression_pointer` to handle one
indexing operation out into its own function,
`BlockContext::write_access_chain_index`.
2024-10-08 11:53:15 -07:00
Jim Blandy 3d85781f05 [naga spv-out] Consolidate code to find index values.
Let the SPIR-V backend use `GuardedIndex::try_resolve_to_constant`,
rather than writing out its definition in `write_restricted_index` and
`write_index_comparison`.

Call `try_resolve_to_constant` in one place, in `write_bounds_check`,
and simply pass the `GuardedIndex` into subroutines.

Reduce `write_restricted_index` and `write_index_comparison` to case
analysis and code generation.

Note that this commit does have a benign effect on SPIR-V snapshot
output for programs like this:

    let one_i = 1i;
    var vec0 = vec3<i32>();
    vec0[one_i] = 1;

The value indexing `vec0` here is an `i32`, but after this commit, the
operand to `OpAccessChain` becomes a `u32` constant (with the same
value).

This is because `write_bounds_check` now calls
`try_resolve_to_constant` itself, rather than deferring this work to
its callees, so it may return `BoundsCheckResult::KnownInBounds` even
when the `Unchecked` policy is in force. This directs the caller,
`write_expression_pointer`, to treat the `OpAccessChain` operand as a
fresh `u32` constant, rather than simply passing through the original
`i32` expression.
2024-10-08 11:53:15 -07:00
Jim Blandy 287ca16b52 [naga spv-out] Abstract extending a bounds check condition chain.
Introduce a new function,
`BlockContext::extend_bounds_check_condition_chain`, which adds a new
boolean condition to the chain of bounds checks guarding an
`OpAccessChain` instruction.
2024-10-08 11:53:15 -07:00
Jim Blandy 21c527a458 [naga spv-out] Doc fix: typo 2024-10-08 11:53:15 -07:00
Jim Blandy 634a97fcb8 [naga spv-out] Abstract out non-uniform binding array access test.
Introduce a new function,
`BlockContext::is_nonuniform_binding_array_access`, which determines
whether a given array access expression means that the `OpAccessChain`
instruction must have a `NonUniform` decoration.
2024-10-08 11:53:15 -07:00
ChosenName 43cb730d58 [naga] added DrawID (#6325) 2024-10-08 13:00:00 +00:00
Jim Blandy e432980a73 [naga spv-out] Don't emit unreachable blocks that jump into loops.
When generating SPIR-V, avoid generating unreachable blocks following
statements like `break`, `return`, and so on that cause non-local
exits. These unreachable blocks can cause SPIR-V validation to fail.

Fixes #6220.
2024-10-03 21:07:18 -07:00
Jim Blandy 04182c24ec [naga spv-out] Make write_block and its types private.
Make `naga::back::spv::BlockContext::write_block` private to
`naga::back::spv::block`. Introduce a new `pub(super)` function,
`write_function_body`, for `Writer::write_function` to call.

Make `BlockExit` private to `naga::back::spv::block`.

Move `LoopContext` from `naga::back::spv` into
`naga::back::spv::block`, and make it private.
2024-10-03 21:07:18 -07:00
Jim Blandy d3e09dd63a [naga spv-out] Consolidate explanation of global wrapping.
Consolidate the explanation of why and how we wrap Naga global
variables in structs to satisfy Vulkan's requirements, and include it
in the documentation for `back::spv::GlobalVariable`.

Clarify `GlobalVariable`'s members documentation.
2024-10-03 10:13:08 -07:00
Erich Gubler 6e528aaf24 style: split (most) string literals over 100-column line width limit 2024-10-03 10:57:28 -04:00
Erich Gubler 9f275f7655 style: remove comments from MathFunction type res. that stop rustfmt 2024-10-03 10:57:28 -04:00
Erich Gubler 25b2b6afa1 style: newline b/w Span::{UNDEFINED, new} 2024-10-03 10:57:28 -04:00
Erich Gubler 78928654a2 style(wgsl-in): unblock rustfmt in Error::as_parse_error 2024-10-03 10:57:28 -04:00
Erich Gubler 3d584f99ed refactor(spv-out): linkify docs. ref. to write_bounds_check 2024-10-02 22:35:09 -04:00
Jim Blandy 215f0fc887 [naga spv-out] Internal doc fixes for back::spv::index.
In `naga::back::spv::index`, clarify documentation for:
- `BoundsCheckResult`
- `write_index_comparison`
- `write_bounds_check`
2024-10-02 22:35:09 -04:00
2d82054ae4 [wgsl-in, spv-out] Allow dynamic indexing of arrays by value.
Bring https://github.com/gfx-rs/naga/pull/723 back from the dead.

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
Co-authored-by: Jim Blandy <jimb@red-bean.com>
2024-10-02 18:07:02 -07:00
Jim Blandy 2021e7f29f [naga spv-out] Document and refactor write_runtime_array_length.
Document and refactor
`naga::back::spv::BlockContext::write_runtime_array_length`.

Don't try to handle finding the length of a particular element of a
`binding_array<array<T>>`. The SPIR-V backend doesn't wrap that type
correctly anyway; #6333 changes the validator to forbid such types.
Instead, assume that the elements of a `binding_array<T>` are always
structs whose final members may be a runtime-sized array.

Pull out consistency checks after the analysis of the array
expression, so that we always carry out all the checks regardless of
what path we took to produce the information.
2024-09-27 17:02:34 -07:00
Jim Blandy 259592b926 [naga spv-out] Update Vulkan spec section number, and provide link. 2024-09-27 17:02:34 -07:00