Produce less noise in ASTs for a common case.
Also test that an empty continuing construct doesn't show up in the
AST. That's currently handled by the AST code. We want to keep this
behaviour even if the AST implementation changes. Right now
code change is needed when emitting the start of a continuing
construct.
Bug: tint:3
Change-Id: I96a12087e305c64647561f65d87acda907ae9c42
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/22844
Reviewed-by: dan sinclair <dsinclair@google.com>
This emits the equivalent of break-if, break-unless, continue-if,
continue-unless. But we do it via a regular if-then-else.
Adds a test matrix.
Adds all required tests except for those needing OpSwitch.
Bug: tint:3
Change-Id: I960a40aa00f95f394a92a099c8b12104010ad49f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/22603
Reviewed-by: dan sinclair <dsinclair@google.com>
A merge block must be dominated by its own header.
This CL checks the cases where that fails because the
merge block is also the:
- the default or case header for a switch (a different header)
- the true-head, false-head, or premerge-head for an if-selection
with a different header
Bug: tint:3
Change-Id: I6dd1fae162e9d33bd9a0b43d3ca9558cebed058b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/22680
Reviewed-by: dan sinclair <dsinclair@google.com>
This fixes the pathological cases nobody wants, and arguably
should be added to the SPIR-V spec.
If we really really want to support these cases, we can revisit.
Bug: tint:3
Change-Id: I0a75490d451676caa0933e3761098ba1fe3f8b60
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/22664
Reviewed-by: dan sinclair <dsinclair@google.com>
The `>>>` symbol was folded into the `>>` symbol in WGSL. This CL
removes `>>>` from Tint.
Change-Id: I9d900de9a6026a8099796b94aad44483f0c6813f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/22582
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
This is preparation for emitting nested control flow.
Bug: tint:3
Change-Id: I90fc7edba8cb9937f722e6f5e94c7f222d34c403
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21801
Reviewed-by: dan sinclair <dsinclair@google.com>
The NOP command was added for complete SPIR-V bijectivity which is no
longer a goal of WGSL. The NOP command has been removed from the spec,
so remove from Tint.
Change-Id: Ic640d9d3b987a19668dc169ff549444921c5bbb8
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/22160
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Do so systmatically. Before we had tested some as a side effect
of other objectives.
Fix the error message for when we have a bad exit from a loop construct
that bypasses not only the continue construct but the loop merge block
itself.
Bug: tint:3
Change-Id: Iaf8fc9bcd3162002aa906efa90a244ef5f439911
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21580
Reviewed-by: dan sinclair <dsinclair@google.com>
Remove TOOD about special code for finding badly nested selections.
This is already adequately covered by the dominance-violation checks,
and related tests:
ClassifyCFGEdges_DomViolation_BeforeIfToSelectionInterior
ClassifyCFGEdges_DomViolation_BeforeSwitchToSelectionInterior
ClassifyCFGEdges_DomViolation_BeforeLoopToLoopBodyInterior
ClassifyCFGEdges_DomViolation_BeforeContinueToContinueInterior
ClassifyCFGEdges_DomViolation_AfterContinueToContinueInterior
Bug: tint:3
Change-Id: I8c547474d1422a54078d748b39a7d22a500d5d1d
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21482
Reviewed-by: dan sinclair <dsinclair@google.com>
Unordered float compares are not supported directly by WGSL.
Translate them as negated ordered compares.
Bug: tint:3
Change-Id: I4fea7c924054cffc9a39a8be3b3d9f088d302114
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21540
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Finds the "then", the "else", and "premerge" nodes.
The premerge node, if it exists, is the first block where
the normal forward flow of the "then" and "else" clauses
converge, but before the merge block.
Finds error case where there a block has both an if-break
edge and a forward-to-premerge. There is no good way
to model that in a high level language.
Bug: tint:3
Change-Id: I759fc539f3480e38d091041db6a9abd15f3df769
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21240
Reviewed-by: dan sinclair <dsinclair@google.com>
This also refactors break detection.
Bug: tint:3
Change-Id: I3a3e01c8d76d7c6fc2a14b3dbff136acd487e802
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21220
Reviewed-by: dan sinclair <dsinclair@google.com>
These used to be classified as kToMerge
Bug: tint:3
Change-Id: I63fb286c371582676e5943a3082c0614b9c8d5d0
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21081
Reviewed-by: dan sinclair <dsinclair@google.com>
Distinguish between selections constructs starting with with
OpBranchConditional and those starting with OpSwitch.
We'll use this in a followup CL to track break from a switch.
Bug: tint:3
Change-Id: I8d000cb42325535a4937c84f83a83c98a9b8d4c5
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/21080
Reviewed-by: dan sinclair <dsinclair@google.com>
Classify CFG edges:
- loop backedge
- a structured exit:
- loop break
- loop continue
- selection break
- fallthrough
- forward (any of the rest)
Also error out when there should have been a merge instruction.
(More than one unique fallthrough or forward edge).
Includes lots of tests.
Bug: tint:3
Change-Id: I70f27680bdf098213056522abf04ac58a6b478ab
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20481
Reviewed-by: dan sinclair <dsinclair@google.com>
This CL splits the TypeConstructor code out of the generic
ConstructorExpression to make the code separation clearer.
Bug: tint:5
Change-Id: I441955c3b09a30b2d02f542f4b7c8e80ff8d9d86
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20640
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
This CL adds the beginning of break and continue support. The
conditional versions are not supported, just the non-conditional.
Bug: tint:5
Change-Id: I84418cffd3e29dc011c4313bf9aa3da4833c009f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20500
Reviewed-by: David Neto <dneto@google.com>
Label basic blocks with:
- their nearest enclosing structured control flow constructs.
- their nearest enclosing continue construct, if any
- their nearest enclosing loop construct, if any
A construct consists of a span of blocks in the computed block order.
It knows its parent construct, if any, and its nesting depth.
Bug: tint:3
Change-Id: Ia945706e8ea2435d6c40fb4e36dc2daeeb9780d0
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20421
Reviewed-by: dan sinclair <dsinclair@google.com>
This is gives us the fundamental ordering of blocks in relation
to a structured construct.
Bug: tint:3
Change-Id: I76eb39403131305398808c33ce4cee256a1c23c2
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20266
Reviewed-by: dan sinclair <dsinclair@google.com>
TEST=Built fuzzer in local checkout of Chromium with tint integrated.
BUG=dawn:14
Change-Id: I84aacd41f893070c3af40f5e640361e177689f9e
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20343
Reviewed-by: dan sinclair <dsinclair@google.com>
Record header/merge cross-links, and single_block_loop attribute of BlockInfo.
Also checks that they are sane: only target blocks in the same function.
Bug: tint:3
Change-Id: I715f7ed354a556e92d58a4c9ba6f306c746c3641
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20080
Reviewed-by: dan sinclair <dsinclair@google.com>
Each basic block must have a terminator, and if the terminator branches
to another block, that block must name a label in the same function.
Bug: tint:3
Change-Id: If22bece6a8041fef362c02b05e4dfee999a3e5bf
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20046
Reviewed-by: dan sinclair <dsinclair@google.com>
Test non-nested sequences and selections.
Bug: tint:3
Change-Id: Ibbbcd428d701d9e7d4da1682f94c2bdbef00121b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/19920
Reviewed-by: dan sinclair <dsinclair@google.com>
Fix bad merge around emission of OpBitcast.
Change-Id: Iacc5ba1c3940745e4b2c46878bbc416397ec555f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20047
Reviewed-by: David Neto <dneto@google.com>
Also, it's not clear if OpSRem has a direct mapping.
Bug: tint:3
Change-Id: Ie7834253cf14109fbebd2ece8e18d9899b29753b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/19881
Reviewed-by: dan sinclair <dsinclair@google.com>
(I expect that) the WGSL signed division operator expects both operands
to be signed and the result will also be signed.
When the operands of a SPIR-V OpSDiv is unsigned, then wrap
the operand in an as-cast to the corresponding signed type.
When the result type of a SPIR-V OpSDiv instruction is unsigned,
we have to wrap the generated WGSL operator with an as-cast to
that unsigned type.
This first CL addresses OpSDiv. We'll address other operations in future CLs.
Bug: tint:3
Change-Id: If3849ceb44b21db87c1efd2c6a2cd63c6d648c88
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/19800
Reviewed-by: dan sinclair <dsinclair@google.com>
The AST only wants expressions, not their result types.
But the SPIR-V reader wants to track the AST type as well.
So introduce a TypedExpression concept for internal use.
Bug: tint:3
Change-Id: Ia832f7422440ef0e8e04630cdca98cae20e18921
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/20040
Reviewed-by: dan sinclair <dsinclair@google.com>
