Apparently, the parameter index calculation in #37861 was not always correct in presence of other pullback parameters (e.g. captures and non-differentiated args). Collect all inout parameters and collect inout parameter adjoints correctly.
Resolves SR-15891
AutoDiff/compiler_crashers_fixed/rdar71191415-nested-differentiation-of-extension-method-optimized.swift is starting to fail recently. Tracked by rdar://89055298.
- Introduce a workaround while dealing with getLoweredParameterIndices() results:
it operates on AST and therefore does not take into account captured arguments
which are "on side" on AST and explicit in SIL
- Ensure captured arguments (@inout_aliased) are handled in a same way as ordinary
@inout arguments while generating a pullback type
Fixes SR-15205
This allows automatic differentiation of concrete calls to math functions operating on `Double`. Such calls were not differentiable because the concrete math functions were imported from C, not the generic ones defined in Swift. The fix is to define derivatives for these C functions. `test/AutoDiff/stdlib/tgmath_derivatives.swift.gyb` was modified to test the new functions.
The "regular" CMO is done with the option `-cross-module-optimization`. It's good for performance but can increase code size.
Now, which this change CMO is also done if the option is not given, but in a very conservative way. Only very small functions are serialized and not additional type metadata is kept alive.
rdar://70082202
Due to rdar://87429620, test/AutoDiff/SILOptimizer/differentiation_diagnostics.swift is still using `-requirement-machine=off`. This patch moves the reproducer to a standalone XFAIL test, and removes `-requirement-machine=off` from differentiation_diagnostics.swift.
Add new `-print-ast-decl` frontend option for only printing declarations,
to match existing behavior.
Some tests want to print the AST, but don't care about expressions.
The existing `-print-ast` option now prints function bodies and expressions.
Not all expressions are printed yet, but most common ones are.
In `TangentBuilder` temporary allocations are emitted in order to make generic calls to `AdditiveArithmetic.zero` and `AdditiveArithmetic.+`. When we perform accumulation of two adjoint values that correspond to a lexical value in the original function, the stack allocations require a debug variable, as we use the original instruction's location. However such a debug variable wouldn't really make sense as these temporary allocations do not represent the original variable's adjoint. With lexical borrow scopes, we hit a crasher when creating temporary allocations when accumulating adjoints for lexical `begin_borrow` instructions.
This PR makes `TangentBuilder`'s temporary allocations use an auto-generated location. The existing debug info emission behavior is not changed, as we still emit `debug_value` when materializing an `AdjointValue` after any necessary adjoint accumulation.
The effect of passing -enable-copy-propagation is both to enable the
CopyPropagation pass to shorten object lifetimes and also to enable
lexical lifetimes to ensure that object lifetimes aren't shortened while
a variable is still in scope and used.
Add a new flag, -enable-lexical-borrow-scopes=true to override
-enable-copy-propagation's effect (setting it to ::ExperimentalLate) on
SILOptions::LexicalLifetimes that sets it to ::Early even in the face of
-enable-copy-propagation. The old flag -disable-lexical-lifetimes is
renamed to -enable-lexical-borrow-scopes=false but continues to set that
option to ::Off even when -enable-copy-propagation is passed.
When checking the viability of an original function candidate as specified in a `@derivative` attribute, a candidate's signautre can have more generic requirements than the required signature. Such cases need to be checked and diagnosed.
Resolves SR-15530 / rdar://85845512.
The latest Long Term Support NDK finally removed binutils, including the bfd/gold
linkers and libgcc. This simplifies our Android support, including making lld the
default linker for Android. Disable three reflection tests that now fail, likely
related to issues with swift-reflection-dump and switching to lld.
This change separates out the formation of the generic signature and
substitutions for a SIL substituted function type as a pre-pass
before doing the actual function type lowering. The only input we
really need to form this signature is the original abstraction pattern
that a type is being lowered against, and pre-computing it should make
the code less side-effecty and confusing. It also allows us to handle
generic nominal types in a more robust way; we transfer over all of
the nominal type requirements to the generalized generic signature,
then when recursively visiting the bindings, we same-type-constrain
the generic parameters used in those requirements to the newly-generalized
generic arguments. This ensures that the minimized signature preserves
any non-trivial requirements imposed by the nominal type, such as
conditional conformances on its type arguments, same-type constraints
among associated types, etc.
This approach does lead to less-than-optimal generalized generic
signatures getting generated, since nominal type generic arguments
get same-type-bound either to other generic arguments or fixed to
concrete types almost always. It would be useful to do a minimization
pass on the final generic signature to eliminate these unnecessary
generic arguments, but that can be done in a follow-up PR.
When SILGenWitnessTable creates a decl ref for the witness of a derivative function requirement, it is using the requirement's derivative generic signature in the resulting witness decl ref. This is wrong because the witness may have a different derivative generic signature than the requirement, leading to a crash. This bug was never discovered because GSB's dark magic made it "just work", until requirement machine.
The fix is to store the matched witness derivative generic signature in `Witness` during type checking, and during witness table generation, use the witness' generic signature to create a witness decl ref.
Resolves rdar://84716758, rdar://84213107 and rdar://84987079.
The latest Long Term Support NDK finally removed binutils, including the bfd/gold
linkers and libgcc. This simplifies our Android support, including making lld the
default linker for Android. Disable three reflection tests that now fail, likely
related to issues with swift-reflection-dump and switching to lld.
Also, add the libatomic dependency for Android armv7, just as on linux.
1. When calculating the differential type of an original function with an inout parameter and when the inout parameter has a type parameter, the inout parameter should get a generic parameter in the subst generic signature of the differential but it currently doesn't. This causes SILGen to attempt to reabstract the differential value in the JVP protocol witness thunk, whilst the generic signature is lacking requirements, leading to a requirement machine error. This patch fixes the calculation so that the JVP's result type (the differential type) always matches the witness thunk's result type.
Wrong type:
```swift
sil private [transparent] [thunk] [ossa] @... <τ_0_0 where τ_0_0 : Differentiable> (...) -> @owned @callee_guaranteed @substituted <τ_0_0, τ_0_1> (@in_guaranteed τ_0_0) -> @out τ_0_1 for <τ_0_0.TangentVector, τ_0_0.TangentVector> {
%6 = differentiable_function_extract [jvp] %5 : $@differentiable(reverse) @convention(method) <τ_0_0 where τ_0_0 : Differentiable> (@in_guaranteed τ_0_0, @noDerivative @inout τ_0_0, @noDerivative SR_13305_Struct) -> () // user: %7
HERE ====> %7 = apply %6<τ_0_0>(%0, %1, %3) : $@convention(method) <τ_0_0 where τ_0_0 : Differentiable> (@in_guaranteed τ_0_0, @inout τ_0_0, SR_13305_Struct) -> @owned @callee_guaranteed @substituted <τ_0_0> (@in_guaranteed τ_0_0) -> @out τ_0_0 for <τ_0_0.TangentVector>
```
Should be:
```swift
%7 = apply %6<τ_0_0>(%0, %1, %3) : $@convention(method) <τ_0_0 where τ_0_0 : Differentiable> (@in_guaranteed τ_0_0, @inout τ_0_0, SR_13305_Struct) -> @owned @callee_guaranteed @substituted <τ_0_0, τ_0_1> (@in_guaranteed τ_0_0) -> @out τ_0_1 for <τ_0_0.TangentVector, τ_0_0.TangentVector>
```
2. `TypeConverter::makeConstantInterfaceType` is not passing down the derivative generic signature to `SILFunctionType::getAutoDiffDerivativeFunctionType` for class methods, and this was caught by RequirementMachine during vtable emission. This patch fixes that.
Partially resolves rdar://82549134. The only remaining tests that require `-requirement-machine=off` are SILOptimizer/semantic_member_accessors_sil.swift and SILOptimizer/differentiation_diagnostics.swift which I will fix next. Then I'll do a proper fix for workaround #39416.
Provides reproducer lit test case for AutoDiff implicit auxiliary struct and enum
declarations that clone implicit GenericTypeParamDecls from a generic
signature into a flat GenericParamList, and lead to a compiler assert about GenericTypeParamDecl depth during merge module operation.
Related to https://github.com/apple/swift/pull/32343.
The SIL type lowering logic for AutoDiff gets the substituted generic signature
mixed up with the invocation generic signature, so it tries to ask questions
about DependentMemberTypes in a signature with no requirements. This triggers
assertions when the requirement machine is enabled.
Disable the requirement machine until this is fixed.
