Auto merge of rust-lang#123244 - Mark-Simulacrum:share-inline-never-generics, r=saethlin

Enable -Zshare-generics for inline(never) functions

This avoids inlining cross-crate generic items when possible that are
already marked inline(never), implying that the author is not intending
for the function to be inlined by callers. As such, having a local copy
may make it easier for LLVM to optimize but mostly just adds to binary
bloat and codegen time. In practice our benchmarks indicate this is
indeed a win for larger compilations, where the extra cost in dynamic
linking to these symbols is diminished compared to the advantages in
fewer copies that need optimizing in each binary.

It might also make sense it expand this with other heuristics (e.g.,
`#[cold]`) in the future, but this seems like a good starting point.

FWIW, I expect that doing cleanup in where we make the decision
what should/shouldn't be shared is also a good idea. Way too
much code needed to be tweaked to check this. But I'm hoping
to leave that for a follow-up PR rather than blocking this on it.

Author: bors

alloc/src/raw_vec.rs

@@ -757,7 +757,9 @@ impl<A: Allocator> RawVecInner<A> {
     }
 }
 
-#[inline(never)]
+// not marked inline(never) since we want optimizers to be able to observe the specifics of this
+// function, see tests/codegen/vec-reserve-extend.rs.
+#[cold]
 fn finish_grow<A>(
     new_layout: Layout,
     current_memory: Option<(NonNull<u8>, Layout)>,

std/src/lib.rs

@@ -362,6 +362,7 @@
 #![feature(strict_provenance_atomic_ptr)]
 #![feature(sync_unsafe_cell)]
 #![feature(ub_checks)]
+#![feature(used_with_arg)]
 // tidy-alphabetical-end
 //
 // Library features (alloc):

std/src/panicking.rs

@@ -27,6 +27,22 @@ use crate::sys::backtrace;
 use crate::sys::stdio::panic_output;
 use crate::{fmt, intrinsics, process, thread};
 
+// This forces codegen of the function called by panic!() inside the std crate, rather than in
+// downstream crates. Primarily this is useful for rustc's codegen tests, which rely on noticing
+// complete removal of panic from generated IR. Since begin_panic is inline(never), it's only
+// codegen'd once per crate-graph so this pushes that to std rather than our codegen test crates.
+//
+// (See https://github.com/rust-lang/rust/pull/123244 for more info on why).
+//
+// If this is causing problems we can also modify those codegen tests to use a crate type like
+// cdylib which doesn't export "Rust" symbols to downstream linkage units.
+#[unstable(feature = "libstd_sys_internals", reason = "used by the panic! macro", issue = "none")]
+#[doc(hidden)]
+#[allow(dead_code)]
+#[used(compiler)]
+pub static EMPTY_PANIC: fn(&'static str) -> ! =
+    begin_panic::<&'static str> as fn(&'static str) -> !;
+
 // Binary interface to the panic runtime that the standard library depends on.
 //
 // The standard library is tagged with `#![needs_panic_runtime]` (introduced in