Rename `assert_uninit_valid` intrinsic
It's not about "uninit" anymore but about "filling with 0x01 bytes" so the name should at least try to reflect that.
This is actually not fully correct though, as it does still panic for all uninit with `-Zstrict-init-checks`. I'm not sure what the best way is to deal with that not causing confusion. I guess we could just remove the flag? I don't think having it makes a lot of sense anymore with the direction that we have chose to go. It could be relevant again if #100423 lands so removing it may be a bit over eager.
r? `@RalfJung`
Use struct types during codegen in less places
This makes it easier to use cg_ssa from a backend like Cranelift that doesn't have any struct types at all. After this PR struct types are still used for function arguments and return values. Removing those usages is harder but should still be doable.
Fix invalid codegen during debuginfo lowering
In order for LLVM to correctly generate debuginfo for msvc, we sometimes need to spill arguments to the stack and perform some direct & indirect offsets into the value. Previously, this code always performed those actions, even when not required as LLVM would clean it up during optimization.
However, when MIR inlining is enabled, this can cause problems as the operations occur prior to the spilled value being initialized. To solve this, we first calculate the necessary offsets using just the type which is side-effect free and does not alter the LLVM IR. Then, if we are in a situation which requires us to generate the LLVM IR (and this situation only occurs for arguments, not local variables) then we perform the same calculation again, this time generating the appropriate LLVM IR as we go.
r? `@tmiasko` but feel free to reassign if you want 🙂Fixes#105386
In order for LLVM to correctly generate debuginfo for msvc, we sometimes
need to spill arguments to the stack and perform some direct & indirect
offsets into the value. Previously, this code always performed those
actions, even when not required as LLVM would clean it up during
optimization.
However, when MIR inlining is enabled, this can cause problems as the
operations occur prior to the spilled value being initialized. To solve
this, we first calculate the necessary offsets using just the type which
is side-effect free and does not alter the LLVM IR. Then, if we are in a
situation which requires us to generate the LLVM IR (and this situation
only occurs for arguments, not local variables) then we perform the same
calculation again, this time generating the appropriate LLVM IR as we
go.
rustc_codegen_ssa: Fix for codegen_get_discr
When doing the optimized implementation of getting the discriminant, the arithmetic needs to be done in the tag type so wrapping behavior works correctly.
Fixes#104519
When doing the optimized implementation of getting the discriminant, the
arithmetic needs to be done in the tag type so wrapping behavior works
correctly.
Fixes#104519
In `codegen_assert_terminator` we decide if a BB's successor is a
candidate for merging, which requires that it be the only successor, and
that it only have one predecessor. That result then gets passed down,
and if it reaches `funclet_br` with the appropriate BB characteristics,
then no `br` instruction is issued, a `MergingSucc::True` result is
passed back, and the merging proceeds in `codegen_block`.
The commit also adds `CachedLlbb`, a new type to help keep track of
each BB that has been merged into its predecessor.
For the next commit, `FunctionCx::codegen_*_terminator` need to take a
`&mut Bx` instead of consuming a `Bx`. This triggers a cascade of
similar changes across multiple functions. The resulting code is more
concise and replaces many `&mut bx` expressions with `bx`.
`codegen_switchint_terminator` already uses `br` instead of `switch`
when there is one normal target plus the `otherwise` target. But there's
another common case with two normal targets and an `otherwise` target
that points to an empty unreachable BB. This comes up a lot when
switching on the tags of enums that use niches.
The pattern looks like this:
```
bb1: ; preds = %bb6
%3 = load i8, ptr %_2, align 1, !range !9, !noundef !4
%4 = sub i8 %3, 2
%5 = icmp eq i8 %4, 0
%_6 = select i1 %5, i64 0, i64 1
switch i64 %_6, label %bb3 [
i64 0, label %bb4
i64 1, label %bb2
]
bb3: ; preds = %bb1
unreachable
```
This commit adds code to convert the `switch` to a `br`:
```
bb1: ; preds = %bb6
%3 = load i8, ptr %_2, align 1, !range !9, !noundef !4
%4 = sub i8 %3, 2
%5 = icmp eq i8 %4, 0
%_6 = select i1 %5, i64 0, i64 1
%6 = icmp eq i64 %_6, 0
br i1 %6, label %bb4, label %bb2
bb3: ; No predecessors!
unreachable
```
This has a surprisingly large effect on compile times, with reductions
of 5% on debug builds of some crates. The reduction is all due to LLVM
taking less time. Maybe LLVM is just much better at handling `br` than
`switch`.
The resulting code is still suboptimal.
- The `icmp`, `select`, `icmp` sequence is silly, converting an `i1` to an `i64`
and back to an `i1`. But with the current code structure it's hard to avoid,
and LLVM will easily clean it up, in opt builds at least.
- `bb3` is usually now truly dead code (though not always, so it can't
be removed universally).
`TerminatorCodegenHelper` has three methods `llblock`, `llbb`, and
`lltarget`. They're all similar, but the names given no indication of
the differences.
This commit renames `lltarget` as `llbb_with_landing_pad`, and `llblock`
as `llbb_with_cleanup`. These aren't fantastic names, but at least it's
now clear that `llbb` is the lowest-level of the three and the other two
wrap it.
Ensure:
- builders always have a `bx` suffix;
- backend basic blocks always have an `llbb` suffix,
- paired builders and basic blocks have consistent prefixes.
Support casting boxes to dyn*
Boxes have a pointer type at codegen time which LLVM does not allow to be transparently converted to an integer. Work around this by inserting a `ptrtoint` instruction if the argument is a pointer.
r? ``@compiler-errors``
Fixes#102427
Initial implementation of dyn*
This PR adds extremely basic and incomplete support for [dyn*](https://smallcultfollowing.com/babysteps//blog/2022/03/29/dyn-can-we-make-dyn-sized/). The goal is to get something in tree behind a flag to make collaboration easier, and also to make sure the implementation so far is not unreasonable. This PR does quite a few things:
* Introduce `dyn_star` feature flag
* Adds parsing for `dyn* Trait` types
* Defines `dyn* Trait` as a sized type
* Adds support for explicit casts, like `42usize as dyn* Debug`
* Including const evaluation of such casts
* Adds codegen for drop glue so things are cleaned up properly when a `dyn* Trait` object goes out of scope
* Adds codegen for method calls, at least for methods that take `&self`
Quite a bit is still missing, but this gives us a starting point. Note that this is never intended to become stable surface syntax for Rust, but rather `dyn*` is planned to be used as an implementation detail for async functions in dyn traits.
Joint work with `@nikomatsakis` and `@compiler-errors.`
r? `@bjorn3`
