codegen: Fix va_list - aarch64 iOS/Windows
## Summary
Fix code generated for `VaList` on Aarch64 iOS/Windows.
## Details
According to the [Apple - ARM64 Function Calling Conventions]:
> ... the type va_list is an alias for char * rather than for the struct
> type specified in the generic PCS.
The current implementation uses the generic Aarch64 structure for `VaList`
for Aarch64 iOS. Switch to using the `char *` variant of the `VaList`
and use the corresponding `emit_ptr_va_arg` for the `va_arg` intrinsic.
Windows always uses the `char *` variant of the `VaList`. Update the `va_arg`
intrinsic to use `emit_ptr_va_arg`.
[Apple - ARM64 Function Calling Conventions]: https://developer.apple.com/library/archive/documentation/Xcode/Conceptual/iPhoneOSABIReference/Articles/ARM64FunctionCallingConventions.html
According to the Apple developer docs:
> The type va_list is an alias for char * rather than for the struct
> type specified in the generic PCS.
The current implementation uses the generic Aarch64 structure for VaList
for Aarch64 iOS.
Windows always uses the char * variant of the va_list.
codegen_llvm_back: improve allocations
This commit was split out from https://github.com/rust-lang/rust/pull/54864. Last time it was causing an LLVM OOM, which was most probably caused by not collecting the globals.
- preallocate vectors of known length
- `extend` instead of `append` where the argument is consumable
- turn 2 `push` loops into `extend`s
- create a vector from a function producing one instead of using `extend_from_slice` on it
- consume `modules` when no longer needed
- ~~return an `impl Iterator` from `generate_lto_work`~~
- ~~don't `collect` `globals`, as they are iterated over and consumed right afterwards~~
While I'm hoping it won't cause an OOM anymore, I would still consider this a "high-risk" PR and not roll it up.
Enable -mergefunc-use-aliases
If the Rust LLVM fork is used, enable the -mergefunc-use-aliases
flag, which will create aliases for merged functions, rather than
inserting a call from one to the other.
A number of codegen tests needed to be adjusted, because functions
that previously fell below the thunk limit are now being merged.
Merging is prevented in various ways now.
I expect that this is going to break something, somewhere, because
it isn't able to deal with aliases properly, but we won't find out
until we try :)
This fixes#52651.
r? @rkruppe
name-anon-globals should always be run at the very end of the pass
pipeline, as optimization passes (in particular mergefunc) may
introduce new anonymous globals.
I believe we did not run into this earlier because it requires the
rather specific combination of a) mergefunc merging two weak functions
b) compilation not using thinlto.
If the Rust LLVM fork is used, enable the -mergefunc-use-aliases
flag, which will create aliases for merged functions, rather than
inserting a call from one to the other.
A number of codegen tests needed to be adjusted, because functions
that previously fell below the thunk limit are now being merged.
Merging is prevented either using -C no-prepopulate-passes, or by
making the functions non-identical.
I expect that this is going to break something, somewhere, because
it isn't able to deal with aliases properly, but we won't find out
until we try :)
This fixes#52651.
libcore: Add VaList and variadic arg handling intrinsics
## Summary
- Add intrinsics for `va_start`, `va_end`, `va_copy`, and `va_arg`.
- Add `core::va_list::VaList` to `libcore`.
Part 1 of (at least) 3 for #44930
Comments and critiques are very much welcomed 😄
Encode a custom "producers" section in wasm files
This commit implements WebAssembly/tool-conventions#65 for wasm files
produced by the Rust compiler. This adds a bit of metadata to wasm
modules to indicate that the file's language includes Rust and the
file's "processed-by" tools includes rustc.
The thinking with this section is to eventually have telemetry in
browsers tracking all this.
Pass additional linker flags when targeting Fuchsia
This is a follow up to 8aa9267 which changed the driver to use lld
directly rather than invoking it through Clang. This change ensures
we pass all the necessary flags to lld.
rustc_target: separate out an individual alignment quantity type from Align.
Before this PR, `rustc_target::abi::Align` combined "power-of-two alignment quantity" semantics, with a distinction between ABI (required) and preferred alignment (by having two quantities).
After this PR, `Align` is only *one* such quantity, and a new `AbiAndPrefAlign` type is introduced to hold the pair of ABI and preferred `Align` quantities.
`Align` is used everywhere one quantity is necessary/sufficient, simplifying some of the code in codegen/miri, while `AbiAndPrefAlign` only in layout computation (to propagate preferred alignment).
r? @oli-obk cc @nagisa @RalfJung @nikomatsakis
This is a follow up to 8aa9267 which changed the driver to use lld
directly rather than invoking it through Clang. This change ensures
we pass all the necessary flags to lld.
This commit implements WebAssembly/tool-conventions#65 for wasm files
produced by the Rust compiler. This adds a bit of metadata to wasm
modules to indicate that the file's language includes Rust and the
file's "processed-by" tools includes rustc.
The thinking with this section is to eventually have telemetry in
browsers tracking all this.
