Auto merge of #45225 - eddyb:trans-abi, r=arielb1

Refactor type memory layouts and ABIs, to be more general and easier to optimize.

To combat combinatorial explosion, type layouts are now described through 3 orthogonal properties:
* `Variants` describes the plurality of sum types (where applicable)
  * `Single` is for one inhabited/active variant, including all C `struct`s and `union`s
  * `Tagged` has its variants discriminated by an integer tag, including C `enum`s
  * `NicheFilling` uses otherwise-invalid values ("niches") for all but one of its inhabited variants
* `FieldPlacement` describes the number and memory offsets of fields (if any)
  * `Union` has all its fields at offset `0`
  * `Array` has offsets that are a multiple of its `stride`; guarantees all fields have one type
  * `Arbitrary` records all the field offsets, which can be out-of-order
* `Abi` describes how values of the type should be passed around, including for FFI
  * `Uninhabited` corresponds to no values, associated with unreachable control-flow
  * `Scalar` is ABI-identical to its only integer/floating-point/pointer "scalar component"
  * `ScalarPair` has two "scalar components", but only applies to the Rust ABI
  * `Vector` is for SIMD vectors, typically `#[repr(simd)]` `struct`s in Rust
  * `Aggregate` has arbitrary contents, including all non-transparent C `struct`s and `union`s

Size optimizations implemented so far:
* ignoring uninhabited variants (i.e. containing uninhabited fields), e.g.:
  * `Option<!>` is 0 bytes
  * `Result<T, !>` has the same size as `T`
* using arbitrary niches, not just `0`, to represent a data-less variant, e.g.:
  * `Option<bool>`, `Option<Option<bool>>`, `Option<Ordering>` are all 1 byte
  * `Option<char>` is 4 bytes
* using a range of niches to represent *multiple* data-less variants, e.g.:
  * `enum E { A(bool), B, C, D }` is 1 byte

Code generation now takes advantage of `Scalar` and `ScalarPair` to, in more cases, pass around scalar components as immediates instead of indirectly, through pointers into temporary memory, while avoiding LLVM's "first-class aggregates", and there's more untapped potential here.

Closes #44426, fixes #5977, fixes #14540, fixes #43278.

1 files changed, 13 insertions, 12 deletions

diff --git a/src/test/codegen/function-arguments.rs b/src/test/codegen/function-arguments.rs
index 29e2840c881..f8945a6ee8d 100644
--- a/src/test/codegen/function-arguments.rs
+++ b/src/test/codegen/function-arguments.rs
@@ -9,12 +9,13 @@
 // except according to those terms.
 
 // compile-flags: -C no-prepopulate-passes
+// ignore-tidy-linelength
 
 #![crate_type = "lib"]
 #![feature(custom_attribute)]
 
 pub struct S {
-  _field: [i64; 4],
+  _field: [i32; 8],
 }
 
 pub struct UnsafeInner {
@@ -45,13 +46,13 @@ pub fn static_borrow(_: &'static i32) {
 pub fn named_borrow<'r>(_: &'r i32) {
 }
 
-// CHECK: @unsafe_borrow(%UnsafeInner* dereferenceable(2) %arg0)
+// CHECK: @unsafe_borrow(i16* dereferenceable(2) %arg0)
 // unsafe interior means this isn't actually readonly and there may be aliases ...
 #[no_mangle]
 pub fn unsafe_borrow(_: &UnsafeInner) {
 }
 
-// CHECK: @mutable_unsafe_borrow(%UnsafeInner* dereferenceable(2) %arg0)
+// CHECK: @mutable_unsafe_borrow(i16* dereferenceable(2) %arg0)
 // ... unless this is a mutable borrow, those never alias
 // ... except that there's this LLVM bug that forces us to not use noalias, see #29485
 #[no_mangle]
@@ -76,7 +77,7 @@ pub fn indirect_struct(_: S) {
 pub fn borrowed_struct(_: &S) {
 }
 
-// CHECK: noalias dereferenceable(4) i32* @_box(i32* noalias dereferenceable(4) %x)
+// CHECK: noalias align 4 dereferenceable(4) i32* @_box(i32* noalias dereferenceable(4) %x)
 #[no_mangle]
 pub fn _box(x: Box<i32>) -> Box<i32> {
   x
@@ -86,7 +87,7 @@ pub fn _box(x: Box<i32>) -> Box<i32> {
 #[no_mangle]
 pub fn struct_return() -> S {
   S {
-    _field: [0, 0, 0, 0]
+    _field: [0, 0, 0, 0, 0, 0, 0, 0]
   }
 }
 
@@ -96,43 +97,43 @@ pub fn struct_return() -> S {
 pub fn helper(_: usize) {
 }
 
-// CHECK: @slice(i8* noalias nonnull readonly %arg0.ptr, [[USIZE]] %arg0.meta)
+// CHECK: @slice([0 x i8]* noalias nonnull readonly %arg0.0, [[USIZE]] %arg0.1)
 // FIXME #25759 This should also have `nocapture`
 #[no_mangle]
 pub fn slice(_: &[u8]) {
 }
 
-// CHECK: @mutable_slice(i8* nonnull %arg0.ptr, [[USIZE]] %arg0.meta)
+// CHECK: @mutable_slice([0 x i8]* nonnull %arg0.0, [[USIZE]] %arg0.1)
 // FIXME #25759 This should also have `nocapture`
 // ... there's this LLVM bug that forces us to not use noalias, see #29485
 #[no_mangle]
 pub fn mutable_slice(_: &mut [u8]) {
 }
 
-// CHECK: @unsafe_slice(%UnsafeInner* nonnull %arg0.ptr, [[USIZE]] %arg0.meta)
+// CHECK: @unsafe_slice([0 x i16]* nonnull %arg0.0, [[USIZE]] %arg0.1)
 // unsafe interior means this isn't actually readonly and there may be aliases ...
 #[no_mangle]
 pub fn unsafe_slice(_: &[UnsafeInner]) {
 }
 
-// CHECK: @str(i8* noalias nonnull readonly %arg0.ptr, [[USIZE]] %arg0.meta)
+// CHECK: @str([0 x i8]* noalias nonnull readonly %arg0.0, [[USIZE]] %arg0.1)
 // FIXME #25759 This should also have `nocapture`
 #[no_mangle]
 pub fn str(_: &[u8]) {
 }
 
-// CHECK: @trait_borrow({}* nonnull, {}* noalias nonnull readonly)
+// CHECK: @trait_borrow(%"core::ops::drop::Drop"* nonnull %arg0.0, {}* noalias nonnull readonly %arg0.1)
 // FIXME #25759 This should also have `nocapture`
 #[no_mangle]
 pub fn trait_borrow(_: &Drop) {
 }
 
-// CHECK: @trait_box({}* noalias nonnull, {}* noalias nonnull readonly)
+// CHECK: @trait_box(%"core::ops::drop::Drop"* noalias nonnull, {}* noalias nonnull readonly)
 #[no_mangle]
 pub fn trait_box(_: Box<Drop>) {
 }
 
-// CHECK: { i16*, [[USIZE]] } @return_slice(i16* noalias nonnull readonly %x.ptr, [[USIZE]] %x.meta)
+// CHECK: { [0 x i16]*, [[USIZE]] } @return_slice([0 x i16]* noalias nonnull readonly %x.0, [[USIZE]] %x.1)
 #[no_mangle]
 pub fn return_slice(x: &[u16]) -> &[u16] {
   x