Auto merge of #111768 - oli-obk:pair_const_llvm, r=cjgillot

Optimize scalar and scalar pair representations loaded from ByRef in llvm

in https://github.com/rust-lang/rust/pull/105653 I noticed that we were generating suboptimal LLVM IR if we had a `ConstValue::ByRef` that could be represented by a `ScalarPair`. Before https://github.com/rust-lang/rust/pull/105653 this is probably rare, but after it, every slice will go down this suboptimal code path that requires LLVM to untangle a bunch of indirections and translate static allocations that are only used once to read a scalar pair from.

1 files changed, 8 insertions, 0 deletions

diff --git a/tests/codegen/const_scalar_pair.rs b/tests/codegen/const_scalar_pair.rs
new file mode 100644
index 00000000000..8f32c50b798
--- /dev/null
+++ b/tests/codegen/const_scalar_pair.rs
@@ -0,0 +1,8 @@
+// compile-flags: --crate-type=lib -Copt-level=0 -Zmir-opt-level=0 -C debuginfo=2
+
+#![feature(inline_const)]
+
+pub fn foo() -> (i32, i32) {
+    // CHECK: ret { i32, i32 } { i32 1, i32 2 }
+    const { (1, 2) }
+}