WebAssembly does not support _Float16 type, so we need to guard the use
of the type. Unfortunately, Clang does not provide a good way to detect
the support of _Float16 type at compile time, so just disable for wasm
targets.
We can easily test all 2**16 values, so let's do it. Also now _Float16 is properly supported in clang, so we can pass arguments to CPP that way, which lets us get snan right on more platforms.
This introduces a non-Darwin (non-CrashReporter) storage for error
messages to allow extraction for crash reporting. This is initially
meant to be used on Windows, though it is generic enough to be used on
any platform.
This didn't expose any missing imports in the sources of any of the standard
library modules, but that's not entirely surprising. Still, we should have the
feature enabled to test it and prevent regressions from creeping in before the
behavior becomes the default.
Exclude `ExtensionImportVisibility` from module interfaces.
The descriptor and arguments for prespecialized metadata will always be in the shared cache. Skip creating the mangling for any lookup involving pointers outside the shared cache, as an optimization.
This PR implements first set of changes required to support autodiff for coroutines. It mostly targeted to `_modify` accessors in standard library (and beyond), but overall implementation is quite generic.
There are some specifics of implementation and known limitations:
- Only `@yield_once` coroutines are naturally supported
- VJP is a coroutine itself: it yields the results *and* returns a pullback closure as a normal return. This allows us to capture values produced in resume part of a coroutine (this is required for defers and other cleanups / commits)
- Pullback is a coroutine, we assume that coroutine cannot abort and therefore we execute the original coroutine in reverse from return via yield and then back to the entry
- It seems there is no semantically sane way to support `_read` coroutines (as we will need to "accept" adjoints via yields), therefore only coroutines with inout yields are supported (`_modify` accessors). Pullbacks of such coroutines take adjoint buffer as input argument, yield this buffer (to accumulate adjoint values in the caller) and finally return the adjoints indirectly.
- Coroutines (as opposed to normal functions) are not first-class values: there is no AST type for them, one cannot e.g. store them into tuples, etc. So, everywhere where AST type is required, we have to hack around.
- As there is no AST type for coroutines, there is no way one could register custom derivative for coroutines. So far only compiler-produced derivatives are supported
- There are lots of common things wrt normal function apply's, but still there are subtle but important differences. I tried to organize the code to enable code reuse, still it was not always possible, so some code duplication could be seen
- The order of how pullback closures are produced in VJP is a bit different: for normal apply's VJP produces both value and pullback closure via a single nested VJP apply. This is not so anymore with coroutine VJP's: yielded values are produced at `begin_apply` site and pullback closure is available only from `end_apply`, so we need to track the order in which pullbacks are produced (and arrange consumption of the values accordingly – effectively delay them)
- On the way some complementary changes were required in e.g. mangler / demangler
This patch covers the generation of derivatives up to SIL level, however, it is not enough as codegen of `partial_apply` of a coroutine is completely broken. The fix for this will be submitted separately as it is not directly autodiff-related.
---------
Co-authored-by: Andrew Savonichev <andrew.savonichev@gmail.com>
Co-authored-by: Richard Wei <rxwei@apple.com>
We need to check for overridden images on every image load, otherwise
XCTest (among others) may `dlopen()` an image that pulls in something
that is overridden, at which point the prespecialized metadata won't
match the image we loaded.
rdar://125727356
* Break ambiguity for types conforming to both UEC and SVEC
If an encoding container conforms to _both_ UEC and SVEC and does not
itself implement support for [U]Int128, there is an ambiguity between
the two default implementations. Add additional defaults defined on
the intersection of the protocols to resolve this.
* Update abi tests for new ambiguity breakers.
The model for associated types hasn't been fully worked-out for
noncopyable generics, but there is some support already that is being
used by the stdlib for an internal-only (and rather cursed) protocol
`_Pointer` to support `UnsafePointer`, etc.
This patch gates the existing experimental support for associated types
behind a feature flag. This flag doesn't emit feature-guards in
interfaces, since support for it is tied closely to NoncopyableGenerics
and has been there from its early days.
Emit metadata for runtime checks of conformances of associated types to
invertible protocols, e.g., `T.Assoc: Copyable`. This allows us to
correctly handle, e.g., dynamic casting involving conditional
conformances that have such constraints.
The model we use here is to emit an invertible-protocol constraint
that leaves only the specific bit clear in the invertible protocol
set.
I had to change the APIs to always be always emit into client instead of back
deployable since silgen_name seems to interfere with @backDeployment. So I
switched the implementation so that it instead uses an always emit into client
thunk with the in source function name and a usableFromInline function that has
the silgen_name. This ensures that we still appropriately export the same symbol
as we did before, so it is ABI stable.
This was approved as part of se-0414.
rdar://122030520
* Revert "Revert count(where:)"
This reverts commit 779ea19a6a.
Now that SE-0220 has been re-accepted, this adds the `count(where:)`
Sequence method to the standard library.
New async iterators should be able to implement only `next(isolation:)` and
get the older `next()` implementation via a default. Implement the
appropriate default witness.
Fixes rdar://125447861.
If an expression refers to noncopyable storage, then default to performing
a borrowing switch, where `let` bindings in patterns borrow out of the
matched value. If an expression refers to a temporary value or explicitly
uses the `consume` keyword, then perform a consuming switch, where
`let` bindings take ownership of corresponding parts of the matched value.
Allow `_borrowing` to still be used to explicitly bind a pattern variable
as a borrow, with no-implicit-copy semantics for copyable values.
