Auto merge of #116270 - cjgillot:gvn-aggregate, r=oli-obk,RalfJung

See through aggregates in GVN

This PR is extracted from https://github.com/rust-lang/rust/pull/111344

The first 2 commit are cleanups to avoid repeated work. I propose to stop removing useless assignments as part of this pass, and let a later `SimplifyLocals` do it. This makes tests easier to read (among others).

The next 3 commits add a constant folding mechanism to the GVN pass, presented in https://github.com/rust-lang/rust/pull/116012. ~This pass is designed to only use global allocations, to avoid any risk of accidental modification of the stored state.~

The following commits implement opportunistic simplifications, in particular:
- projections of aggregates: `MyStruct { x: a }.x` gets replaced by `a`, works with enums too;
- projections of arrays: `[a, b][0]` becomes `a`;
- projections of repeat expressions: `[a; N][x]` becomes `a`;
- transform arrays of equal operands into a repeat rvalue.

Fixes https://github.com/rust-lang/miri/issues/3090

r? `@oli-obk`

compiler/rustc_middle/src/mir/consts.rs

@@ -172,6 +172,24 @@ impl<'tcx> ConstValue<'tcx> {
         let end = end.try_into().unwrap();
         Some(data.inner().inspect_with_uninit_and_ptr_outside_interpreter(start..end))
     }
+
+    /// Check if a constant may contain provenance information. This is used by MIR opts.
+    /// Can return `true` even if there is no provenance.
+    pub fn may_have_provenance(&self, tcx: TyCtxt<'tcx>, size: Size) -> bool {
+        match *self {
+            ConstValue::ZeroSized | ConstValue::Scalar(Scalar::Int(_)) => return false,
+            ConstValue::Scalar(Scalar::Ptr(..)) => return true,
+            // It's hard to find out the part of the allocation we point to;
+            // just conservatively check everything.
+            ConstValue::Slice { data, meta: _ } => !data.inner().provenance().ptrs().is_empty(),
+            ConstValue::Indirect { alloc_id, offset } => !tcx
+                .global_alloc(alloc_id)
+                .unwrap_memory()
+                .inner()
+                .provenance()
+                .range_empty(super::AllocRange::from(offset..offset + size), &tcx),
+        }
+    }
 }
 
 ///////////////////////////////////////////////////////////////////////////
@@ -485,6 +503,38 @@ impl<'tcx> Const<'tcx> {
             _ => Self::Ty(c),
         }
     }
+
+    /// Return true if any evaluation of this constant always returns the same value,
+    /// taking into account even pointer identity tests.
+    pub fn is_deterministic(&self) -> bool {
+        // Some constants may generate fresh allocations for pointers they contain,
+        // so using the same constant twice can yield two different results:
+        // - valtrees purposefully generate new allocations
+        // - ConstValue::Slice also generate new allocations
+        match self {
+            Const::Ty(c) => match c.kind() {
+                ty::ConstKind::Param(..) => true,
+                // A valtree may be a reference. Valtree references correspond to a
+                // different allocation each time they are evaluated. Valtrees for primitive
+                // types are fine though.
+                ty::ConstKind::Value(_) => c.ty().is_primitive(),
+                ty::ConstKind::Unevaluated(..) | ty::ConstKind::Expr(..) => false,
+                // Should not appear in runtime MIR.
+                ty::ConstKind::Infer(..)
+                | ty::ConstKind::Bound(..)
+                | ty::ConstKind::Placeholder(..)
+                | ty::ConstKind::Error(..) => bug!(),
+            },
+            Const::Unevaluated(..) => false,
+            // If the same slice appears twice in the MIR, we cannot guarantee that we will
+            // give the same `AllocId` to the data.
+            Const::Val(ConstValue::Slice { .. }, _) => false,
+            Const::Val(
+                ConstValue::ZeroSized | ConstValue::Scalar(_) | ConstValue::Indirect { .. },
+                _,
+            ) => true,
+        }
+    }
 }
 
 /// An unevaluated (potentially generic) constant used in MIR.