Merge from rust-lang/rust

18 files changed, 332 insertions, 295 deletions

diff --git a/compiler/rustc_codegen_ssa/Cargo.toml b/compiler/rustc_codegen_ssa/Cargo.toml
index 2afc1efc1b3..83d847f2d15 100644
--- a/compiler/rustc_codegen_ssa/Cargo.toml
+++ b/compiler/rustc_codegen_ssa/Cargo.toml
@@ -1,7 +1,7 @@
 [package]
 name = "rustc_codegen_ssa"
 version = "0.0.0"
-edition = "2021"
+edition = "2024"
 
 [dependencies]
 # tidy-alphabetical-start
@@ -25,6 +25,7 @@ rustc_data_structures = { path = "../rustc_data_structures" }
 rustc_errors = { path = "../rustc_errors" }
 rustc_fluent_macro = { path = "../rustc_fluent_macro" }
 rustc_fs_util = { path = "../rustc_fs_util" }
+rustc_hashes = { path = "../rustc_hashes" }
 rustc_hir = { path = "../rustc_hir" }
 rustc_hir_pretty = { path = "../rustc_hir_pretty" }
 rustc_incremental = { path = "../rustc_incremental" }
diff --git a/compiler/rustc_codegen_ssa/src/back/link.rs b/compiler/rustc_codegen_ssa/src/back/link.rs
index 105a4cb81f0..4c076450269 100644
--- a/compiler/rustc_codegen_ssa/src/back/link.rs
+++ b/compiler/rustc_codegen_ssa/src/back/link.rs
@@ -244,22 +244,17 @@ pub fn each_linked_rlib(
 
         fmts
     } else {
-        for combination in info.dependency_formats.iter().combinations(2) {
-            let (ty1, list1) = &combination[0];
-            let (ty2, list2) = &combination[1];
-            if list1 != list2 {
-                return Err(errors::LinkRlibError::IncompatibleDependencyFormats {
-                    ty1: format!("{ty1:?}"),
-                    ty2: format!("{ty2:?}"),
-                    list1: format!("{list1:?}"),
-                    list2: format!("{list2:?}"),
-                });
-            }
-        }
-        if info.dependency_formats.is_empty() {
-            return Err(errors::LinkRlibError::MissingFormat);
+        let mut dep_formats = info.dependency_formats.iter();
+        let (ty1, list1) = dep_formats.next().ok_or(errors::LinkRlibError::MissingFormat)?;
+        if let Some((ty2, list2)) = dep_formats.find(|(_, list2)| list1 != *list2) {
+            return Err(errors::LinkRlibError::IncompatibleDependencyFormats {
+                ty1: format!("{ty1:?}"),
+                ty2: format!("{ty2:?}"),
+                list1: format!("{list1:?}"),
+                list2: format!("{list2:?}"),
+            });
         }
-        info.dependency_formats.first().unwrap().1
+        list1
     };
 
     let used_dep_crates = info.used_crates.iter();
@@ -626,10 +621,9 @@ fn link_staticlib(
 
     let mut all_rust_dylibs = vec![];
     for &cnum in crates {
-        match fmts.get(cnum) {
-            Some(&Linkage::Dynamic) => {}
-            _ => continue,
-        }
+        let Some(Linkage::Dynamic) = fmts.get(cnum) else {
+            continue;
+        };
         let crate_name = codegen_results.crate_info.crate_name[&cnum];
         let used_crate_source = &codegen_results.crate_info.used_crate_source[&cnum];
         if let Some((path, _)) = &used_crate_source.dylib {
@@ -1990,6 +1984,7 @@ fn add_pre_link_args(cmd: &mut dyn Linker, sess: &Session, flavor: LinkerFlavor)
     if let Some(args) = sess.target.pre_link_args.get(&flavor) {
         cmd.verbatim_args(args.iter().map(Deref::deref));
     }
+
     cmd.verbatim_args(&sess.opts.unstable_opts.pre_link_args);
 }
 
@@ -2518,6 +2513,12 @@ fn add_order_independent_options(
             "--target-cpu",
             &codegen_results.crate_info.target_cpu,
         ]);
+        if codegen_results.crate_info.target_features.len() > 0 {
+            cmd.link_arg(&format!(
+                "--target-feature={}",
+                &codegen_results.crate_info.target_features.join(",")
+            ));
+        }
     } else if flavor == LinkerFlavor::Ptx {
         cmd.link_args(&["--fallback-arch", &codegen_results.crate_info.target_cpu]);
     } else if flavor == LinkerFlavor::Bpf {
diff --git a/compiler/rustc_codegen_ssa/src/back/linker.rs b/compiler/rustc_codegen_ssa/src/back/linker.rs
index e3ace01c1eb..8900405c1b8 100644
--- a/compiler/rustc_codegen_ssa/src/back/linker.rs
+++ b/compiler/rustc_codegen_ssa/src/back/linker.rs
@@ -153,6 +153,7 @@ pub(crate) fn get_linker<'a>(
             hinted_static: None,
             is_ld: cc == Cc::No,
             is_gnu: flavor.is_gnu(),
+            uses_lld: flavor.uses_lld(),
         }) as Box<dyn Linker>,
         LinkerFlavor::Msvc(..) => Box::new(MsvcLinker { cmd, sess }) as Box<dyn Linker>,
         LinkerFlavor::EmCc => Box::new(EmLinker { cmd, sess }) as Box<dyn Linker>,
@@ -361,6 +362,7 @@ struct GccLinker<'a> {
     // Link as ld
     is_ld: bool,
     is_gnu: bool,
+    uses_lld: bool,
 }
 
 impl<'a> GccLinker<'a> {
@@ -552,6 +554,7 @@ impl<'a> Linker for GccLinker<'a> {
                 self.link_args(&["--entry", "_initialize"]);
             }
         }
+
         // VxWorks compiler driver introduced `--static-crt` flag specifically for rustc,
         // it switches linking for libc and similar system libraries to static without using
         // any `#[link]` attributes in the `libc` crate, see #72782 for details.
@@ -567,6 +570,15 @@ impl<'a> Linker for GccLinker<'a> {
         {
             self.cc_arg("--static-crt");
         }
+
+        // avr-none doesn't have default ISA, users must specify which specific
+        // CPU (well, microcontroller) they are targetting using `-Ctarget-cpu`.
+        //
+        // Currently this makes sense only when using avr-gcc as a linker, since
+        // it brings a couple of hand-written important intrinsics from libgcc.
+        if self.sess.target.arch == "avr" && !self.uses_lld {
+            self.verbatim_arg(format!("-mmcu={}", self.target_cpu));
+        }
     }
 
     fn link_dylib_by_name(&mut self, name: &str, verbatim: bool, as_needed: bool) {
diff --git a/compiler/rustc_codegen_ssa/src/back/metadata.rs b/compiler/rustc_codegen_ssa/src/back/metadata.rs
index d70413b8a47..236507ac0cd 100644
--- a/compiler/rustc_codegen_ssa/src/back/metadata.rs
+++ b/compiler/rustc_codegen_ssa/src/back/metadata.rs
@@ -252,15 +252,7 @@ pub(crate) fn create_object_file(sess: &Session) -> Option<write::Object<'static
         // Unsupported architecture.
         _ => return None,
     };
-    let binary_format = if sess.target.is_like_osx {
-        BinaryFormat::MachO
-    } else if sess.target.is_like_windows {
-        BinaryFormat::Coff
-    } else if sess.target.is_like_aix {
-        BinaryFormat::Xcoff
-    } else {
-        BinaryFormat::Elf
-    };
+    let binary_format = sess.target.binary_format.to_object();
 
     let mut file = write::Object::new(binary_format, architecture, endianness);
     file.set_sub_architecture(sub_architecture);
diff --git a/compiler/rustc_codegen_ssa/src/back/symbol_export.rs b/compiler/rustc_codegen_ssa/src/back/symbol_export.rs
index 12ee872d531..459f4329d6e 100644
--- a/compiler/rustc_codegen_ssa/src/back/symbol_export.rs
+++ b/compiler/rustc_codegen_ssa/src/back/symbol_export.rs
@@ -183,11 +183,11 @@ fn exported_symbols_provider_local(
     });
 
     let mut symbols: Vec<_> =
-        sorted.iter().map(|(&def_id, &info)| (ExportedSymbol::NonGeneric(def_id), info)).collect();
+        sorted.iter().map(|&(&def_id, &info)| (ExportedSymbol::NonGeneric(def_id), info)).collect();
 
     // Export TLS shims
     if !tcx.sess.target.dll_tls_export {
-        symbols.extend(sorted.iter().filter_map(|(&def_id, &info)| {
+        symbols.extend(sorted.iter().filter_map(|&(&def_id, &info)| {
             tcx.needs_thread_local_shim(def_id).then(|| {
                 (
                     ExportedSymbol::ThreadLocalShim(def_id),
diff --git a/compiler/rustc_codegen_ssa/src/back/write.rs b/compiler/rustc_codegen_ssa/src/back/write.rs
index f029c08a808..f008bd12ed8 100644
--- a/compiler/rustc_codegen_ssa/src/back/write.rs
+++ b/compiler/rustc_codegen_ssa/src/back/write.rs
@@ -405,7 +405,8 @@ fn generate_lto_work<B: ExtraBackendMethods>(
             B::run_fat_lto(cgcx, needs_fat_lto, import_only_modules).unwrap_or_else(|e| e.raise());
         if cgcx.lto == Lto::Fat && !autodiff.is_empty() {
             let config = cgcx.config(ModuleKind::Regular);
-            module = unsafe { module.autodiff(cgcx, autodiff, config).unwrap() };
+            module =
+                unsafe { module.autodiff(cgcx, autodiff, config).unwrap_or_else(|e| e.raise()) };
         }
         // We are adding a single work item, so the cost doesn't matter.
         vec![(WorkItem::LTO(module), 0)]
@@ -572,10 +573,10 @@ fn produce_final_output_artifacts(
     };
 
     let copy_if_one_unit = |output_type: OutputType, keep_numbered: bool| {
-        if compiled_modules.modules.len() == 1 {
+        if let [module] = &compiled_modules.modules[..] {
             // 1) Only one codegen unit. In this case it's no difficulty
             //    to copy `foo.0.x` to `foo.x`.
-            let module_name = Some(&compiled_modules.modules[0].name[..]);
+            let module_name = Some(&module.name[..]);
             let path = crate_output.temp_path(output_type, module_name);
             let output = crate_output.path(output_type);
             if !output_type.is_text_output() && output.is_tty() {
@@ -707,8 +708,8 @@ fn produce_final_output_artifacts(
     }
 
     if sess.opts.json_artifact_notifications {
-        if compiled_modules.modules.len() == 1 {
-            compiled_modules.modules[0].for_each_output(|_path, ty| {
+        if let [module] = &compiled_modules.modules[..] {
+            module.for_each_output(|_path, ty| {
                 if sess.opts.output_types.contains_key(&ty) {
                     let descr = ty.shorthand();
                     // for single cgu file is renamed to drop cgu specific suffix
@@ -864,7 +865,7 @@ pub(crate) fn compute_per_cgu_lto_type(
     // require LTO so the request for LTO is always unconditionally
     // passed down to the backend, but we don't actually want to do
     // anything about it yet until we've got a final product.
-    let is_rlib = sess_crate_types.len() == 1 && sess_crate_types[0] == CrateType::Rlib;
+    let is_rlib = matches!(sess_crate_types, [CrateType::Rlib]);
 
     match sess_lto {
         Lto::ThinLocal if !linker_does_lto && !is_allocator => ComputedLtoType::Thin,
@@ -1537,8 +1538,9 @@ fn start_executing_work<B: ExtraBackendMethods>(
             // Spin up what work we can, only doing this while we've got available
             // parallelism slots and work left to spawn.
             if codegen_state != Aborted {
-                while !work_items.is_empty() && running_with_own_token < tokens.len() {
-                    let (item, _) = work_items.pop().unwrap();
+                while running_with_own_token < tokens.len()
+                    && let Some((item, _)) = work_items.pop()
+                {
                     spawn_work(
                         &cgcx,
                         &mut llvm_start_time,
diff --git a/compiler/rustc_codegen_ssa/src/base.rs b/compiler/rustc_codegen_ssa/src/base.rs
index e800492dad0..40238f4b491 100644
--- a/compiler/rustc_codegen_ssa/src/base.rs
+++ b/compiler/rustc_codegen_ssa/src/base.rs
@@ -24,7 +24,7 @@ use rustc_middle::query::Providers;
 use rustc_middle::ty::layout::{HasTyCtxt, HasTypingEnv, LayoutOf, TyAndLayout};
 use rustc_middle::ty::{self, Instance, Ty, TyCtxt};
 use rustc_session::Session;
-use rustc_session::config::{self, CrateType, EntryFnType, OptLevel, OutputType};
+use rustc_session::config::{self, CrateType, EntryFnType, OutputType};
 use rustc_span::{DUMMY_SP, Symbol, sym};
 use rustc_trait_selection::infer::{BoundRegionConversionTime, TyCtxtInferExt};
 use rustc_trait_selection::traits::{ObligationCause, ObligationCtxt};
@@ -364,13 +364,7 @@ pub(crate) fn build_shift_expr_rhs<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
     let rhs_sz = bx.cx().int_width(rhs_llty);
     let lhs_sz = bx.cx().int_width(lhs_llty);
     if lhs_sz < rhs_sz {
-        if is_unchecked && bx.sess().opts.optimize != OptLevel::No {
-            // FIXME: Use `trunc nuw` once that's available
-            let inrange = bx.icmp(IntPredicate::IntULE, rhs, mask);
-            bx.assume(inrange);
-        }
-
-        bx.trunc(rhs, lhs_llty)
+        if is_unchecked { bx.unchecked_utrunc(rhs, lhs_llty) } else { bx.trunc(rhs, lhs_llty) }
     } else if lhs_sz > rhs_sz {
         // We zero-extend even if the RHS is signed. So e.g. `(x: i32) << -1i8` will zero-extend the
         // RHS to `255i32`. But then we mask the shift amount to be within the size of the LHS
@@ -921,6 +915,7 @@ impl CrateInfo {
         let n_crates = crates.len();
         let mut info = CrateInfo {
             target_cpu,
+            target_features: tcx.global_backend_features(()).clone(),
             crate_types,
             exported_symbols,
             linked_symbols,
diff --git a/compiler/rustc_codegen_ssa/src/codegen_attrs.rs b/compiler/rustc_codegen_ssa/src/codegen_attrs.rs
index 3e9dfcea58b..97bc84c0b69 100644
--- a/compiler/rustc_codegen_ssa/src/codegen_attrs.rs
+++ b/compiler/rustc_codegen_ssa/src/codegen_attrs.rs
@@ -1,7 +1,6 @@
 use std::str::FromStr;
 
 use rustc_abi::ExternAbi;
-use rustc_ast::attr::list_contains_name;
 use rustc_ast::expand::autodiff_attrs::{
     AutoDiffAttrs, DiffActivity, DiffMode, valid_input_activity, valid_ret_activity,
 };
@@ -377,24 +376,20 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: LocalDefId) -> CodegenFnAttrs {
                             let segments =
                                 set.path.segments.iter().map(|x| x.ident.name).collect::<Vec<_>>();
                             match segments.as_slice() {
-                                [sym::arm, sym::a32] | [sym::arm, sym::t32] => {
-                                    if !tcx.sess.target.has_thumb_interworking {
-                                        struct_span_code_err!(
-                                            tcx.dcx(),
-                                            attr.span,
-                                            E0779,
-                                            "target does not support `#[instruction_set]`"
-                                        )
-                                        .emit();
-                                        None
-                                    } else if segments[1] == sym::a32 {
-                                        Some(InstructionSetAttr::ArmA32)
-                                    } else if segments[1] == sym::t32 {
-                                        Some(InstructionSetAttr::ArmT32)
-                                    } else {
-                                        unreachable!()
-                                    }
+                                [sym::arm, sym::a32 | sym::t32]
+                                    if !tcx.sess.target.has_thumb_interworking =>
+                                {
+                                    struct_span_code_err!(
+                                        tcx.dcx(),
+                                        attr.span,
+                                        E0779,
+                                        "target does not support `#[instruction_set]`"
+                                    )
+                                    .emit();
+                                    None
                                 }
+                                [sym::arm, sym::a32] => Some(InstructionSetAttr::ArmA32),
+                                [sym::arm, sym::t32] => Some(InstructionSetAttr::ArmT32),
                                 _ => {
                                     struct_span_code_err!(
                                         tcx.dcx(),
@@ -435,7 +430,7 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: LocalDefId) -> CodegenFnAttrs {
                     && let Some((sym::align, literal)) = item.singleton_lit_list()
                 {
                     rustc_attr_parsing::parse_alignment(&literal.kind)
-                        .map_err(|msg| {
+                        .inspect_err(|msg| {
                             struct_span_code_err!(
                                 tcx.dcx(),
                                 literal.span,
@@ -536,25 +531,27 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: LocalDefId) -> CodegenFnAttrs {
         }
 
         if attr.is_word() {
-            InlineAttr::Hint
-        } else if let Some(ref items) = attr.meta_item_list() {
-            inline_span = Some(attr.span);
-            if items.len() != 1 {
-                struct_span_code_err!(tcx.dcx(), attr.span, E0534, "expected one argument").emit();
-                InlineAttr::None
-            } else if list_contains_name(items, sym::always) {
-                InlineAttr::Always
-            } else if list_contains_name(items, sym::never) {
-                InlineAttr::Never
-            } else {
-                struct_span_code_err!(tcx.dcx(), items[0].span(), E0535, "invalid argument")
-                    .with_help("valid inline arguments are `always` and `never`")
-                    .emit();
+            return InlineAttr::Hint;
+        }
+        let Some(ref items) = attr.meta_item_list() else {
+            return ia;
+        };
 
-                InlineAttr::None
-            }
+        inline_span = Some(attr.span);
+        let [item] = &items[..] else {
+            struct_span_code_err!(tcx.dcx(), attr.span, E0534, "expected one argument").emit();
+            return InlineAttr::None;
+        };
+        if item.has_name(sym::always) {
+            InlineAttr::Always
+        } else if item.has_name(sym::never) {
+            InlineAttr::Never
         } else {
-            ia
+            struct_span_code_err!(tcx.dcx(), item.span(), E0535, "invalid argument")
+                .with_help("valid inline arguments are `always` and `never`")
+                .emit();
+
+            InlineAttr::None
         }
     });
     codegen_fn_attrs.inline = attrs.iter().fold(codegen_fn_attrs.inline, |ia, attr| {
@@ -586,23 +583,25 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: LocalDefId) -> CodegenFnAttrs {
         let err = |sp, s| struct_span_code_err!(tcx.dcx(), sp, E0722, "{}", s).emit();
         if attr.is_word() {
             err(attr.span, "expected one argument");
-            ia
-        } else if let Some(ref items) = attr.meta_item_list() {
-            inline_span = Some(attr.span);
-            if items.len() != 1 {
-                err(attr.span, "expected one argument");
-                OptimizeAttr::Default
-            } else if list_contains_name(items, sym::size) {
-                OptimizeAttr::Size
-            } else if list_contains_name(items, sym::speed) {
-                OptimizeAttr::Speed
-            } else if list_contains_name(items, sym::none) {
-                OptimizeAttr::DoNotOptimize
-            } else {
-                err(items[0].span(), "invalid argument");
-                OptimizeAttr::Default
-            }
+            return ia;
+        }
+        let Some(ref items) = attr.meta_item_list() else {
+            return OptimizeAttr::Default;
+        };
+
+        inline_span = Some(attr.span);
+        let [item] = &items[..] else {
+            err(attr.span, "expected one argument");
+            return OptimizeAttr::Default;
+        };
+        if item.has_name(sym::size) {
+            OptimizeAttr::Size
+        } else if item.has_name(sym::speed) {
+            OptimizeAttr::Speed
+        } else if item.has_name(sym::none) {
+            OptimizeAttr::DoNotOptimize
         } else {
+            err(item.span(), "invalid argument");
             OptimizeAttr::Default
         }
     });
@@ -644,25 +643,20 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: LocalDefId) -> CodegenFnAttrs {
     // llvm/llvm-project#70563).
     if !codegen_fn_attrs.target_features.is_empty()
         && matches!(codegen_fn_attrs.inline, InlineAttr::Always)
+        && let Some(span) = inline_span
     {
-        if let Some(span) = inline_span {
-            tcx.dcx().span_err(span, "cannot use `#[inline(always)]` with `#[target_feature]`");
-        }
+        tcx.dcx().span_err(span, "cannot use `#[inline(always)]` with `#[target_feature]`");
     }
 
-    if !codegen_fn_attrs.no_sanitize.is_empty() && codegen_fn_attrs.inline.always() {
-        if let (Some(no_sanitize_span), Some(inline_span)) = (no_sanitize_span, inline_span) {
-            let hir_id = tcx.local_def_id_to_hir_id(did);
-            tcx.node_span_lint(
-                lint::builtin::INLINE_NO_SANITIZE,
-                hir_id,
-                no_sanitize_span,
-                |lint| {
-                    lint.primary_message("`no_sanitize` will have no effect after inlining");
-                    lint.span_note(inline_span, "inlining requested here");
-                },
-            )
-        }
+    if !codegen_fn_attrs.no_sanitize.is_empty()
+        && codegen_fn_attrs.inline.always()
+        && let (Some(no_sanitize_span), Some(inline_span)) = (no_sanitize_span, inline_span)
+    {
+        let hir_id = tcx.local_def_id_to_hir_id(did);
+        tcx.node_span_lint(lint::builtin::INLINE_NO_SANITIZE, hir_id, no_sanitize_span, |lint| {
+            lint.primary_message("`no_sanitize` will have no effect after inlining");
+            lint.span_note(inline_span, "inlining requested here");
+        })
     }
 
     // Weak lang items have the same semantics as "std internal" symbols in the
@@ -692,10 +686,10 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: LocalDefId) -> CodegenFnAttrs {
     // Any linkage to LLVM intrinsics for now forcibly marks them all as never
     // unwinds since LLVM sometimes can't handle codegen which `invoke`s
     // intrinsic functions.
-    if let Some(name) = &codegen_fn_attrs.link_name {
-        if name.as_str().starts_with("llvm.") {
-            codegen_fn_attrs.flags |= CodegenFnAttrFlags::NEVER_UNWIND;
-        }
+    if let Some(name) = &codegen_fn_attrs.link_name
+        && name.as_str().starts_with("llvm.")
+    {
+        codegen_fn_attrs.flags |= CodegenFnAttrFlags::NEVER_UNWIND;
     }
 
     if let Some(features) = check_tied_features(
@@ -756,18 +750,13 @@ fn should_inherit_track_caller(tcx: TyCtxt<'_>, def_id: DefId) -> bool {
 
 fn check_link_ordinal(tcx: TyCtxt<'_>, attr: &hir::Attribute) -> Option<u16> {
     use rustc_ast::{LitIntType, LitKind, MetaItemLit};
-    let meta_item_list = attr.meta_item_list();
-    let meta_item_list = meta_item_list.as_deref();
-    let sole_meta_list = match meta_item_list {
-        Some([item]) => item.lit(),
-        Some(_) => {
-            tcx.dcx().emit_err(errors::InvalidLinkOrdinalNargs { span: attr.span });
-            return None;
-        }
-        _ => None,
+    let meta_item_list = attr.meta_item_list()?;
+    let [sole_meta_list] = &meta_item_list[..] else {
+        tcx.dcx().emit_err(errors::InvalidLinkOrdinalNargs { span: attr.span });
+        return None;
     };
     if let Some(MetaItemLit { kind: LitKind::Int(ordinal, LitIntType::Unsuffixed), .. }) =
-        sole_meta_list
+        sole_meta_list.lit()
     {
         // According to the table at
         // https://docs.microsoft.com/en-us/windows/win32/debug/pe-format#import-header, the
@@ -894,7 +883,7 @@ fn autodiff_attrs(tcx: TyCtxt<'_>, id: DefId) -> Option<AutoDiffAttrs> {
     let [mode, input_activities @ .., ret_activity] = &list[..] else {
         span_bug!(attr.span, "rustc_autodiff attribute must contain mode and activities");
     };
-    let mode = if let MetaItemInner::MetaItem(MetaItem { path: ref p1, .. }) = mode {
+    let mode = if let MetaItemInner::MetaItem(MetaItem { path: p1, .. }) = mode {
         p1.segments.first().unwrap().ident
     } else {
         span_bug!(attr.span, "rustc_autodiff attribute must contain mode");
@@ -910,7 +899,7 @@ fn autodiff_attrs(tcx: TyCtxt<'_>, id: DefId) -> Option<AutoDiffAttrs> {
     };
 
     // First read the ret symbol from the attribute
-    let ret_symbol = if let MetaItemInner::MetaItem(MetaItem { path: ref p1, .. }) = ret_activity {
+    let ret_symbol = if let MetaItemInner::MetaItem(MetaItem { path: p1, .. }) = ret_activity {
         p1.segments.first().unwrap().ident
     } else {
         span_bug!(attr.span, "rustc_autodiff attribute must contain the return activity");
@@ -924,7 +913,7 @@ fn autodiff_attrs(tcx: TyCtxt<'_>, id: DefId) -> Option<AutoDiffAttrs> {
     // Now parse all the intermediate (input) activities
     let mut arg_activities: Vec<DiffActivity> = vec![];
     for arg in input_activities {
-        let arg_symbol = if let MetaItemInner::MetaItem(MetaItem { path: ref p2, .. }) = arg {
+        let arg_symbol = if let MetaItemInner::MetaItem(MetaItem { path: p2, .. }) = arg {
             match p2.segments.first() {
                 Some(x) => x.ident,
                 None => {
diff --git a/compiler/rustc_codegen_ssa/src/debuginfo/type_names.rs b/compiler/rustc_codegen_ssa/src/debuginfo/type_names.rs
index 53953b089c6..84703a0a156 100644
--- a/compiler/rustc_codegen_ssa/src/debuginfo/type_names.rs
+++ b/compiler/rustc_codegen_ssa/src/debuginfo/type_names.rs
@@ -15,7 +15,8 @@ use std::fmt::Write;
 
 use rustc_abi::Integer;
 use rustc_data_structures::fx::FxHashSet;
-use rustc_data_structures::stable_hasher::{Hash64, HashStable, StableHasher};
+use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
+use rustc_hashes::Hash64;
 use rustc_hir::def_id::DefId;
 use rustc_hir::definitions::{DefPathData, DefPathDataName, DisambiguatedDefPathData};
 use rustc_hir::{CoroutineDesugaring, CoroutineKind, CoroutineSource, Mutability};
diff --git a/compiler/rustc_codegen_ssa/src/lib.rs b/compiler/rustc_codegen_ssa/src/lib.rs
index 428a45975f1..9d2ac219d59 100644
--- a/compiler/rustc_codegen_ssa/src/lib.rs
+++ b/compiler/rustc_codegen_ssa/src/lib.rs
@@ -190,6 +190,7 @@ impl From<&cstore::NativeLib> for NativeLib {
 #[derive(Debug, Encodable, Decodable)]
 pub struct CrateInfo {
     pub target_cpu: String,
+    pub target_features: Vec<String>,
     pub crate_types: Vec<CrateType>,
     pub exported_symbols: UnordMap<CrateType, Vec<String>>,
     pub linked_symbols: FxIndexMap<CrateType, Vec<(String, SymbolExportKind)>>,
@@ -230,6 +231,7 @@ pub fn provide(providers: &mut Providers) {
     crate::base::provide(providers);
     crate::target_features::provide(providers);
     crate::codegen_attrs::provide(providers);
+    providers.queries.global_backend_features = |_tcx: TyCtxt<'_>, ()| vec![];
 }
 
 /// Checks if the given filename ends with the `.rcgu.o` extension that `rustc`
diff --git a/compiler/rustc_codegen_ssa/src/mir/block.rs b/compiler/rustc_codegen_ssa/src/mir/block.rs
index 4630ed48c52..1e52c639463 100644
--- a/compiler/rustc_codegen_ssa/src/mir/block.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/block.rs
@@ -4,9 +4,7 @@ use rustc_abi::{BackendRepr, ExternAbi, HasDataLayout, Reg, WrappingRange};
 use rustc_ast as ast;
 use rustc_ast::{InlineAsmOptions, InlineAsmTemplatePiece};
 use rustc_hir::lang_items::LangItem;
-use rustc_middle::mir::{
-    self, AssertKind, BasicBlock, InlineAsmMacro, SwitchTargets, UnwindTerminateReason,
-};
+use rustc_middle::mir::{self, AssertKind, InlineAsmMacro, SwitchTargets, UnwindTerminateReason};
 use rustc_middle::ty::layout::{HasTyCtxt, LayoutOf, ValidityRequirement};
 use rustc_middle::ty::print::{with_no_trimmed_paths, with_no_visible_paths};
 use rustc_middle::ty::{self, Instance, Ty};
@@ -429,11 +427,34 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
             let cmp = bx.icmp(IntPredicate::IntEQ, discr_value, llval);
             bx.cond_br(cmp, ll1, ll2);
         } else {
-            bx.switch(
-                discr_value,
-                helper.llbb_with_cleanup(self, targets.otherwise()),
-                target_iter.map(|(value, target)| (value, helper.llbb_with_cleanup(self, target))),
-            );
+            let otherwise = targets.otherwise();
+            let otherwise_cold = self.cold_blocks[otherwise];
+            let otherwise_unreachable = self.mir[otherwise].is_empty_unreachable();
+            let cold_count = targets.iter().filter(|(_, target)| self.cold_blocks[*target]).count();
+            let none_cold = cold_count == 0;
+            let all_cold = cold_count == targets.iter().len();
+            if (none_cold && (!otherwise_cold || otherwise_unreachable))
+                || (all_cold && (otherwise_cold || otherwise_unreachable))
+            {
+                // All targets have the same weight,
+                // or `otherwise` is unreachable and it's the only target with a different weight.
+                bx.switch(
+                    discr_value,
+                    helper.llbb_with_cleanup(self, targets.otherwise()),
+                    target_iter
+                        .map(|(value, target)| (value, helper.llbb_with_cleanup(self, target))),
+                );
+            } else {
+                // Targets have different weights
+                bx.switch_with_weights(
+                    discr_value,
+                    helper.llbb_with_cleanup(self, targets.otherwise()),
+                    otherwise_cold,
+                    target_iter.map(|(value, target)| {
+                        (value, helper.llbb_with_cleanup(self, target), self.cold_blocks[target])
+                    }),
+                );
+            }
         }
     }
 
@@ -699,14 +720,14 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
 
         // Put together the arguments to the panic entry point.
         let (lang_item, args) = match msg {
-            AssertKind::BoundsCheck { ref len, ref index } => {
+            AssertKind::BoundsCheck { len, index } => {
                 let len = self.codegen_operand(bx, len).immediate();
                 let index = self.codegen_operand(bx, index).immediate();
                 // It's `fn panic_bounds_check(index: usize, len: usize)`,
                 // and `#[track_caller]` adds an implicit third argument.
                 (LangItem::PanicBoundsCheck, vec![index, len, location])
             }
-            AssertKind::MisalignedPointerDereference { ref required, ref found } => {
+            AssertKind::MisalignedPointerDereference { required, found } => {
                 let required = self.codegen_operand(bx, required).immediate();
                 let found = self.codegen_operand(bx, found).immediate();
                 // It's `fn panic_misaligned_pointer_dereference(required: usize, found: usize)`,
@@ -919,7 +940,6 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                     &fn_abi.ret,
                     &mut llargs,
                     Some(intrinsic),
-                    target,
                 );
                 let dest = match ret_dest {
                     _ if fn_abi.ret.is_indirect() => llargs[0],
@@ -975,23 +995,17 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
         };
 
         let mut llargs = Vec::with_capacity(arg_count);
-        let destination = target.as_ref().map(|&target| {
-            (
-                self.make_return_dest(
-                    bx,
-                    destination,
-                    &fn_abi.ret,
-                    &mut llargs,
-                    None,
-                    Some(target),
-                ),
-                target,
-            )
-        });
+
+        // We still need to call `make_return_dest` even if there's no `target`, since
+        // `fn_abi.ret` could be `PassMode::Indirect`, even if it is uninhabited,
+        // and `make_return_dest` adds the return-place indirect pointer to `llargs`.
+        let return_dest = self.make_return_dest(bx, destination, &fn_abi.ret, &mut llargs, None);
+        let destination = target.map(|target| (return_dest, target));
 
         // Split the rust-call tupled arguments off.
-        let (first_args, untuple) = if abi == ExternAbi::RustCall && !args.is_empty() {
-            let (tup, args) = args.split_last().unwrap();
+        let (first_args, untuple) = if abi == ExternAbi::RustCall
+            && let Some((tup, args)) = args.split_last()
+        {
             (args, Some(tup))
         } else {
             (args, None)
@@ -1017,7 +1031,7 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                             let (idx, _) = op.layout.non_1zst_field(bx).expect(
                                 "not exactly one non-1-ZST field in a `DispatchFromDyn` type",
                             );
-                            op = op.extract_field(bx, idx);
+                            op = op.extract_field(self, bx, idx);
                         }
 
                         // Now that we have `*dyn Trait` or `&dyn Trait`, split it up into its
@@ -1049,7 +1063,7 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                             let (idx, _) = op.layout.non_1zst_field(bx).expect(
                                 "not exactly one non-1-ZST field in a `DispatchFromDyn` type",
                             );
-                            op = op.extract_field(bx, idx);
+                            op = op.extract_field(self, bx, idx);
                         }
 
                         // Make sure that we've actually unwrapped the rcvr down
@@ -1549,9 +1563,9 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                     if scalar.is_bool() {
                         bx.range_metadata(llval, WrappingRange { start: 0, end: 1 });
                     }
+                    // We store bools as `i8` so we need to truncate to `i1`.
+                    llval = bx.to_immediate_scalar(llval, scalar);
                 }
-                // We store bools as `i8` so we need to truncate to `i1`.
-                llval = bx.to_immediate(llval, arg.layout);
             }
         }
 
@@ -1581,7 +1595,7 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
         } else {
             // If the tuple is immediate, the elements are as well.
             for i in 0..tuple.layout.fields.count() {
-                let op = tuple.extract_field(bx, i);
+                let op = tuple.extract_field(self, bx, i);
                 self.codegen_argument(bx, op, llargs, &args[i]);
             }
         }
@@ -1790,11 +1804,7 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
         fn_ret: &ArgAbi<'tcx, Ty<'tcx>>,
         llargs: &mut Vec<Bx::Value>,
         intrinsic: Option<ty::IntrinsicDef>,
-        target: Option<BasicBlock>,
     ) -> ReturnDest<'tcx, Bx::Value> {
-        if target.is_none() {
-            return ReturnDest::Nothing;
-        }
         // If the return is ignored, we can just return a do-nothing `ReturnDest`.
         if fn_ret.is_ignore() {
             return ReturnDest::Nothing;
diff --git a/compiler/rustc_codegen_ssa/src/mir/mod.rs b/compiler/rustc_codegen_ssa/src/mir/mod.rs
index 3a896071bc6..ba28720afec 100644
--- a/compiler/rustc_codegen_ssa/src/mir/mod.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/mod.rs
@@ -502,14 +502,25 @@ fn find_cold_blocks<'tcx>(
     for (bb, bb_data) in traversal::postorder(mir) {
         let terminator = bb_data.terminator();
 
-        // If a BB ends with a call to a cold function, mark it as cold.
-        if let mir::TerminatorKind::Call { ref func, .. } = terminator.kind
-            && let ty::FnDef(def_id, ..) = *func.ty(local_decls, tcx).kind()
-            && let attrs = tcx.codegen_fn_attrs(def_id)
-            && attrs.flags.contains(CodegenFnAttrFlags::COLD)
-        {
-            cold_blocks[bb] = true;
-            continue;
+        match terminator.kind {
+            // If a BB ends with a call to a cold function, mark it as cold.
+            mir::TerminatorKind::Call { ref func, .. }
+            | mir::TerminatorKind::TailCall { ref func, .. }
+                if let ty::FnDef(def_id, ..) = *func.ty(local_decls, tcx).kind()
+                    && let attrs = tcx.codegen_fn_attrs(def_id)
+                    && attrs.flags.contains(CodegenFnAttrFlags::COLD) =>
+            {
+                cold_blocks[bb] = true;
+                continue;
+            }
+
+            // If a BB ends with an `unreachable`, also mark it as cold.
+            mir::TerminatorKind::Unreachable => {
+                cold_blocks[bb] = true;
+                continue;
+            }
+
+            _ => {}
         }
 
         // If all successors of a BB are cold and there's at least one of them, mark this BB as cold
diff --git a/compiler/rustc_codegen_ssa/src/mir/naked_asm.rs b/compiler/rustc_codegen_ssa/src/mir/naked_asm.rs
index eb0711dbb32..0593fb420c3 100644
--- a/compiler/rustc_codegen_ssa/src/mir/naked_asm.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/naked_asm.rs
@@ -8,7 +8,7 @@ use rustc_middle::ty::{Instance, Ty, TyCtxt};
 use rustc_middle::{bug, span_bug, ty};
 use rustc_span::sym;
 use rustc_target::callconv::{ArgAbi, FnAbi, PassMode};
-use rustc_target::spec::WasmCAbi;
+use rustc_target::spec::{BinaryFormat, WasmCAbi};
 
 use crate::common;
 use crate::traits::{AsmCodegenMethods, BuilderMethods, GlobalAsmOperandRef, MiscCodegenMethods};
@@ -104,27 +104,6 @@ fn inline_to_global_operand<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
     }
 }
 
-enum AsmBinaryFormat {
-    Elf,
-    Macho,
-    Coff,
-    Wasm,
-}
-
-impl AsmBinaryFormat {
-    fn from_target(target: &rustc_target::spec::Target) -> Self {
-        if target.is_like_windows {
-            Self::Coff
-        } else if target.is_like_osx {
-            Self::Macho
-        } else if target.is_like_wasm {
-            Self::Wasm
-        } else {
-            Self::Elf
-        }
-    }
-}
-
 fn prefix_and_suffix<'tcx>(
     tcx: TyCtxt<'tcx>,
     instance: Instance<'tcx>,
@@ -134,7 +113,7 @@ fn prefix_and_suffix<'tcx>(
 ) -> (String, String) {
     use std::fmt::Write;
 
-    let asm_binary_format = AsmBinaryFormat::from_target(&tcx.sess.target);
+    let asm_binary_format = &tcx.sess.target.binary_format;
 
     let is_arm = tcx.sess.target.arch == "arm";
     let is_thumb = tcx.sess.unstable_target_features.contains(&sym::thumb_mode);
@@ -178,10 +157,13 @@ fn prefix_and_suffix<'tcx>(
             }
             Linkage::LinkOnceAny | Linkage::LinkOnceODR | Linkage::WeakAny | Linkage::WeakODR => {
                 match asm_binary_format {
-                    AsmBinaryFormat::Elf | AsmBinaryFormat::Coff | AsmBinaryFormat::Wasm => {
+                    BinaryFormat::Elf
+                    | BinaryFormat::Coff
+                    | BinaryFormat::Wasm
+                    | BinaryFormat::Xcoff => {
                         writeln!(w, ".weak {asm_name}")?;
                     }
-                    AsmBinaryFormat::Macho => {
+                    BinaryFormat::MachO => {
                         writeln!(w, ".globl {asm_name}")?;
                         writeln!(w, ".weak_definition {asm_name}")?;
                     }
@@ -207,7 +189,7 @@ fn prefix_and_suffix<'tcx>(
     let mut begin = String::new();
     let mut end = String::new();
     match asm_binary_format {
-        AsmBinaryFormat::Elf => {
+        BinaryFormat::Elf | BinaryFormat::Xcoff => {
             let section = link_section.unwrap_or(format!(".text.{asm_name}"));
 
             let progbits = match is_arm {
@@ -239,7 +221,7 @@ fn prefix_and_suffix<'tcx>(
                 writeln!(end, "{}", arch_suffix).unwrap();
             }
         }
-        AsmBinaryFormat::Macho => {
+        BinaryFormat::MachO => {
             let section = link_section.unwrap_or("__TEXT,__text".to_string());
             writeln!(begin, ".pushsection {},regular,pure_instructions", section).unwrap();
             writeln!(begin, ".balign {align}").unwrap();
@@ -255,7 +237,7 @@ fn prefix_and_suffix<'tcx>(
                 writeln!(end, "{}", arch_suffix).unwrap();
             }
         }
-        AsmBinaryFormat::Coff => {
+        BinaryFormat::Coff => {
             let section = link_section.unwrap_or(format!(".text.{asm_name}"));
             writeln!(begin, ".pushsection {},\"xr\"", section).unwrap();
             writeln!(begin, ".balign {align}").unwrap();
@@ -272,7 +254,7 @@ fn prefix_and_suffix<'tcx>(
                 writeln!(end, "{}", arch_suffix).unwrap();
             }
         }
-        AsmBinaryFormat::Wasm => {
+        BinaryFormat::Wasm => {
             let section = link_section.unwrap_or(format!(".text.{asm_name}"));
 
             writeln!(begin, ".section {section},\"\",@").unwrap();
diff --git a/compiler/rustc_codegen_ssa/src/mir/operand.rs b/compiler/rustc_codegen_ssa/src/mir/operand.rs
index 9ca7d4f8f00..461cf1b8acd 100644
--- a/compiler/rustc_codegen_ssa/src/mir/operand.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/operand.rs
@@ -1,15 +1,14 @@
-use std::assert_matches::assert_matches;
 use std::fmt;
 
 use arrayvec::ArrayVec;
 use either::Either;
 use rustc_abi as abi;
 use rustc_abi::{Align, BackendRepr, Size};
-use rustc_middle::bug;
 use rustc_middle::mir::interpret::{Pointer, Scalar, alloc_range};
 use rustc_middle::mir::{self, ConstValue};
 use rustc_middle::ty::Ty;
 use rustc_middle::ty::layout::{LayoutOf, TyAndLayout};
+use rustc_middle::{bug, span_bug};
 use tracing::debug;
 
 use super::place::{PlaceRef, PlaceValue};
@@ -352,79 +351,83 @@ impl<'a, 'tcx, V: CodegenObject> OperandRef<'tcx, V> {
 
     pub(crate) fn extract_field<Bx: BuilderMethods<'a, 'tcx, Value = V>>(
         &self,
+        fx: &mut FunctionCx<'a, 'tcx, Bx>,
         bx: &mut Bx,
         i: usize,
     ) -> Self {
         let field = self.layout.field(bx.cx(), i);
         let offset = self.layout.fields.offset(i);
 
-        let mut val = match (self.val, self.layout.backend_repr) {
-            // If the field is ZST, it has no data.
-            _ if field.is_zst() => OperandValue::ZeroSized,
-
-            // Newtype of a scalar, scalar pair or vector.
-            (OperandValue::Immediate(_) | OperandValue::Pair(..), _)
-                if field.size == self.layout.size =>
+        if !bx.is_backend_ref(self.layout) && bx.is_backend_ref(field) {
+            if let BackendRepr::Vector { count, .. } = self.layout.backend_repr
+                && let BackendRepr::Memory { sized: true } = field.backend_repr
+                && count.is_power_of_two()
             {
-                assert_eq!(offset.bytes(), 0);
-                self.val
+                assert_eq!(field.size, self.layout.size);
+                // This is being deprecated, but for now stdarch still needs it for
+                // Newtype vector of array, e.g. #[repr(simd)] struct S([i32; 4]);
+                let place = PlaceRef::alloca(bx, field);
+                self.val.store(bx, place.val.with_type(self.layout));
+                return bx.load_operand(place);
+            } else {
+                // Part of https://github.com/rust-lang/compiler-team/issues/838
+                bug!("Non-ref type {self:?} cannot project to ref field type {field:?}");
             }
+        }
 
-            // Extract a scalar component from a pair.
-            (OperandValue::Pair(a_llval, b_llval), BackendRepr::ScalarPair(a, b)) => {
-                if offset.bytes() == 0 {
-                    assert_eq!(field.size, a.size(bx.cx()));
-                    OperandValue::Immediate(a_llval)
-                } else {
-                    assert_eq!(offset, a.size(bx.cx()).align_to(b.align(bx.cx()).abi));
-                    assert_eq!(field.size, b.size(bx.cx()));
-                    OperandValue::Immediate(b_llval)
+        let val = if field.is_zst() {
+            OperandValue::ZeroSized
+        } else if field.size == self.layout.size {
+            assert_eq!(offset.bytes(), 0);
+            fx.codegen_transmute_operand(bx, *self, field).unwrap_or_else(|| {
+                bug!(
+                    "Expected `codegen_transmute_operand` to handle equal-size \
+                      field {i:?} projection from {self:?} to {field:?}"
+                )
+            })
+        } else {
+            let (in_scalar, imm) = match (self.val, self.layout.backend_repr) {
+                // Extract a scalar component from a pair.
+                (OperandValue::Pair(a_llval, b_llval), BackendRepr::ScalarPair(a, b)) => {
+                    if offset.bytes() == 0 {
+                        assert_eq!(field.size, a.size(bx.cx()));
+                        (Some(a), a_llval)
+                    } else {
+                        assert_eq!(offset, a.size(bx.cx()).align_to(b.align(bx.cx()).abi));
+                        assert_eq!(field.size, b.size(bx.cx()));
+                        (Some(b), b_llval)
+                    }
                 }
-            }
-
-            // `#[repr(simd)]` types are also immediate.
-            (OperandValue::Immediate(llval), BackendRepr::Vector { .. }) => {
-                OperandValue::Immediate(bx.extract_element(llval, bx.cx().const_usize(i as u64)))
-            }
 
-            _ => bug!("OperandRef::extract_field({:?}): not applicable", self),
+                _ => {
+                    span_bug!(fx.mir.span, "OperandRef::extract_field({:?}): not applicable", self)
+                }
+            };
+            OperandValue::Immediate(match field.backend_repr {
+                BackendRepr::Vector { .. } => imm,
+                BackendRepr::Scalar(out_scalar) => {
+                    let Some(in_scalar) = in_scalar else {
+                        span_bug!(
+                            fx.mir.span,
+                            "OperandRef::extract_field({:?}): missing input scalar for output scalar",
+                            self
+                        )
+                    };
+                    if in_scalar != out_scalar {
+                        // If the backend and backend_immediate types might differ,
+                        // flip back to the backend type then to the new immediate.
+                        // This avoids nop truncations, but still handles things like
+                        // Bools in union fields needs to be truncated.
+                        let backend = bx.from_immediate(imm);
+                        bx.to_immediate_scalar(backend, out_scalar)
+                    } else {
+                        imm
+                    }
+                }
+                BackendRepr::ScalarPair(_, _) | BackendRepr::Memory { .. } => bug!(),
+            })
         };
 
-        match (&mut val, field.backend_repr) {
-            (OperandValue::ZeroSized, _) => {}
-            (
-                OperandValue::Immediate(llval),
-                BackendRepr::Scalar(_) | BackendRepr::ScalarPair(..) | BackendRepr::Vector { .. },
-            ) => {
-                // Bools in union fields needs to be truncated.
-                *llval = bx.to_immediate(*llval, field);
-            }
-            (OperandValue::Pair(a, b), BackendRepr::ScalarPair(a_abi, b_abi)) => {
-                // Bools in union fields needs to be truncated.
-                *a = bx.to_immediate_scalar(*a, a_abi);
-                *b = bx.to_immediate_scalar(*b, b_abi);
-            }
-            // Newtype vector of array, e.g. #[repr(simd)] struct S([i32; 4]);
-            (OperandValue::Immediate(llval), BackendRepr::Memory { sized: true }) => {
-                assert_matches!(self.layout.backend_repr, BackendRepr::Vector { .. });
-
-                let llfield_ty = bx.cx().backend_type(field);
-
-                // Can't bitcast an aggregate, so round trip through memory.
-                let llptr = bx.alloca(field.size, field.align.abi);
-                bx.store(*llval, llptr, field.align.abi);
-                *llval = bx.load(llfield_ty, llptr, field.align.abi);
-            }
-            (
-                OperandValue::Immediate(_),
-                BackendRepr::Uninhabited | BackendRepr::Memory { sized: false },
-            ) => {
-                bug!()
-            }
-            (OperandValue::Pair(..), _) => bug!(),
-            (OperandValue::Ref(..), _) => bug!(),
-        }
-
         OperandRef { val, layout: field }
     }
 }
@@ -581,13 +584,13 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                 // Moves out of scalar and scalar pair fields are trivial.
                 for elem in place_ref.projection.iter() {
                     match elem {
-                        mir::ProjectionElem::Field(ref f, _) => {
+                        mir::ProjectionElem::Field(f, _) => {
                             assert!(
                                 !o.layout.ty.is_any_ptr(),
                                 "Bad PlaceRef: destructing pointers should use cast/PtrMetadata, \
                                  but tried to access field {f:?} of pointer {o:?}",
                             );
-                            o = o.extract_field(bx, f.index());
+                            o = o.extract_field(self, bx, f.index());
                         }
                         mir::ProjectionElem::Index(_)
                         | mir::ProjectionElem::ConstantIndex { .. } => {
diff --git a/compiler/rustc_codegen_ssa/src/mir/place.rs b/compiler/rustc_codegen_ssa/src/mir/place.rs
index eb4270ffe80..edd09b9c3c5 100644
--- a/compiler/rustc_codegen_ssa/src/mir/place.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/place.rs
@@ -1,7 +1,7 @@
 use rustc_abi::Primitive::{Int, Pointer};
 use rustc_abi::{Align, BackendRepr, FieldsShape, Size, TagEncoding, VariantIdx, Variants};
+use rustc_middle::mir::PlaceTy;
 use rustc_middle::mir::interpret::Scalar;
-use rustc_middle::mir::tcx::PlaceTy;
 use rustc_middle::ty::layout::{HasTyCtxt, LayoutOf, TyAndLayout};
 use rustc_middle::ty::{self, Ty};
 use rustc_middle::{bug, mir};
@@ -423,7 +423,7 @@ impl<'a, 'tcx, V: CodegenObject> PlaceRef<'tcx, V> {
             layout.size
         };
 
-        let llval = bx.inbounds_gep(bx.cx().backend_type(layout), self.val.llval, &[llindex]);
+        let llval = bx.inbounds_nuw_gep(bx.cx().backend_type(layout), self.val.llval, &[llindex]);
         let align = self.val.align.restrict_for_offset(offset);
         PlaceValue::new_sized(llval, align).with_type(layout)
     }
diff --git a/compiler/rustc_codegen_ssa/src/mir/rvalue.rs b/compiler/rustc_codegen_ssa/src/mir/rvalue.rs
index 4c5b183cfe9..1eebe04225b 100644
--- a/compiler/rustc_codegen_ssa/src/mir/rvalue.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/rvalue.rs
@@ -231,7 +231,7 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
     ///
     /// Returns `None` for cases that can't work in that framework, such as for
     /// `Immediate`->`Ref` that needs an `alloc` to get the location.
-    fn codegen_transmute_operand(
+    pub(crate) fn codegen_transmute_operand(
         &mut self,
         bx: &mut Bx,
         operand: OperandRef<'tcx, Bx::Value>,
@@ -260,6 +260,8 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
             OperandValue::Ref(source_place_val) => {
                 assert_eq!(source_place_val.llextra, None);
                 assert_matches!(operand_kind, OperandValueKind::Ref);
+                // The existing alignment is part of `source_place_val`,
+                // so that alignment will be used, not `cast`'s.
                 Some(bx.load_operand(source_place_val.with_type(cast)).val)
             }
             OperandValue::ZeroSized => {
@@ -664,9 +666,15 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                         lhs.layout.ty,
                     ),
 
-                    (OperandValue::Immediate(lhs_val), OperandValue::Immediate(rhs_val)) => {
-                        self.codegen_scalar_binop(bx, op, lhs_val, rhs_val, lhs.layout.ty)
-                    }
+                    (OperandValue::Immediate(lhs_val), OperandValue::Immediate(rhs_val)) => self
+                        .codegen_scalar_binop(
+                            bx,
+                            op,
+                            lhs_val,
+                            rhs_val,
+                            lhs.layout.ty,
+                            rhs.layout.ty,
+                        ),
 
                     _ => bug!(),
                 };
@@ -887,10 +895,11 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
         op: mir::BinOp,
         lhs: Bx::Value,
         rhs: Bx::Value,
-        input_ty: Ty<'tcx>,
+        lhs_ty: Ty<'tcx>,
+        rhs_ty: Ty<'tcx>,
     ) -> Bx::Value {
-        let is_float = input_ty.is_floating_point();
-        let is_signed = input_ty.is_signed();
+        let is_float = lhs_ty.is_floating_point();
+        let is_signed = lhs_ty.is_signed();
         match op {
             mir::BinOp::Add => {
                 if is_float {
@@ -956,9 +965,9 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
             mir::BinOp::BitAnd => bx.and(lhs, rhs),
             mir::BinOp::BitXor => bx.xor(lhs, rhs),
             mir::BinOp::Offset => {
-                let pointee_type = input_ty
+                let pointee_type = lhs_ty
                     .builtin_deref(true)
-                    .unwrap_or_else(|| bug!("deref of non-pointer {:?}", input_ty));
+                    .unwrap_or_else(|| bug!("deref of non-pointer {:?}", lhs_ty));
                 let pointee_layout = bx.cx().layout_of(pointee_type);
                 if pointee_layout.is_zst() {
                     // `Offset` works in terms of the size of pointee,
@@ -966,7 +975,11 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                     lhs
                 } else {
                     let llty = bx.cx().backend_type(pointee_layout);
-                    bx.inbounds_gep(llty, lhs, &[rhs])
+                    if !rhs_ty.is_signed() {
+                        bx.inbounds_nuw_gep(llty, lhs, &[rhs])
+                    } else {
+                        bx.inbounds_gep(llty, lhs, &[rhs])
+                    }
                 }
             }
             mir::BinOp::Shl | mir::BinOp::ShlUnchecked => {
diff --git a/compiler/rustc_codegen_ssa/src/mono_item.rs b/compiler/rustc_codegen_ssa/src/mono_item.rs
index 5f95b6615bd..f6af889fd6e 100644
--- a/compiler/rustc_codegen_ssa/src/mono_item.rs
+++ b/compiler/rustc_codegen_ssa/src/mono_item.rs
@@ -36,7 +36,7 @@ impl<'a, 'tcx: 'a> MonoItemExt<'a, 'tcx> for MonoItem<'tcx> {
             }
             MonoItem::GlobalAsm(item_id) => {
                 let item = cx.tcx().hir_item(item_id);
-                if let hir::ItemKind::GlobalAsm(asm) = item.kind {
+                if let hir::ItemKind::GlobalAsm { asm, .. } = item.kind {
                     let operands: Vec<_> = asm
                         .operands
                         .iter()
@@ -71,11 +71,8 @@ impl<'a, 'tcx: 'a> MonoItemExt<'a, 'tcx> for MonoItem<'tcx> {
                                     }
                                 }
                             }
-                            hir::InlineAsmOperand::SymFn { ref anon_const } => {
-                                let ty = cx
-                                    .tcx()
-                                    .typeck_body(anon_const.body)
-                                    .node_type(anon_const.hir_id);
+                            hir::InlineAsmOperand::SymFn { expr } => {
+                                let ty = cx.tcx().typeck(item_id.owner_id).expr_ty(expr);
                                 let instance = match ty.kind() {
                                     &ty::FnDef(def_id, args) => Instance::new(def_id, args),
                                     _ => span_bug!(*op_sp, "asm sym is not a function"),
diff --git a/compiler/rustc_codegen_ssa/src/traits/builder.rs b/compiler/rustc_codegen_ssa/src/traits/builder.rs
index 2c843e2f5e4..99fd6b6510f 100644
--- a/compiler/rustc_codegen_ssa/src/traits/builder.rs
+++ b/compiler/rustc_codegen_ssa/src/traits/builder.rs
@@ -1,7 +1,7 @@
 use std::assert_matches::assert_matches;
 use std::ops::Deref;
 
-use rustc_abi::{Align, BackendRepr, Scalar, Size, WrappingRange};
+use rustc_abi::{Align, Scalar, Size, WrappingRange};
 use rustc_middle::middle::codegen_fn_attrs::CodegenFnAttrs;
 use rustc_middle::ty::layout::{FnAbiOf, LayoutOf, TyAndLayout};
 use rustc_middle::ty::{Instance, Ty};
@@ -110,6 +110,20 @@ pub trait BuilderMethods<'a, 'tcx>:
         else_llbb: Self::BasicBlock,
         cases: impl ExactSizeIterator<Item = (u128, Self::BasicBlock)>,
     );
+
+    // This is like `switch()`, but every case has a bool flag indicating whether it's cold.
+    //
+    // Default implementation throws away the cold flags and calls `switch()`.
+    fn switch_with_weights(
+        &mut self,
+        v: Self::Value,
+        else_llbb: Self::BasicBlock,
+        _else_is_cold: bool,
+        cases: impl ExactSizeIterator<Item = (u128, Self::BasicBlock, bool)>,
+    ) {
+        self.switch(v, else_llbb, cases.map(|(val, bb, _)| (val, bb)))
+    }
+
     fn invoke(
         &mut self,
         llty: Self::Type,
@@ -209,13 +223,6 @@ pub trait BuilderMethods<'a, 'tcx>:
     ) -> (Self::Value, Self::Value);
 
     fn from_immediate(&mut self, val: Self::Value) -> Self::Value;
-    fn to_immediate(&mut self, val: Self::Value, layout: TyAndLayout<'_>) -> Self::Value {
-        if let BackendRepr::Scalar(scalar) = layout.backend_repr {
-            self.to_immediate_scalar(val, scalar)
-        } else {
-            val
-        }
-    }
     fn to_immediate_scalar(&mut self, val: Self::Value, scalar: Scalar) -> Self::Value;
 
     fn alloca(&mut self, size: Size, align: Align) -> Self::Value;
@@ -318,6 +325,14 @@ pub trait BuilderMethods<'a, 'tcx>:
         ptr: Self::Value,
         indices: &[Self::Value],
     ) -> Self::Value;
+    fn inbounds_nuw_gep(
+        &mut self,
+        ty: Self::Type,
+        ptr: Self::Value,
+        indices: &[Self::Value],
+    ) -> Self::Value {
+        self.inbounds_gep(ty, ptr, indices)
+    }
     fn ptradd(&mut self, ptr: Self::Value, offset: Self::Value) -> Self::Value {
         self.gep(self.cx().type_i8(), ptr, &[offset])
     }
@@ -326,6 +341,17 @@ pub trait BuilderMethods<'a, 'tcx>:
     }
 
     fn trunc(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value;
+    /// Produces the same value as [`Self::trunc`] (and defaults to that),
+    /// but is UB unless the *zero*-extending the result can reproduce `val`.
+    fn unchecked_utrunc(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value {
+        self.trunc(val, dest_ty)
+    }
+    /// Produces the same value as [`Self::trunc`] (and defaults to that),
+    /// but is UB unless the *sign*-extending the result can reproduce `val`.
+    fn unchecked_strunc(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value {
+        self.trunc(val, dest_ty)
+    }
+
     fn sext(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value;
     fn fptoui_sat(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value;
     fn fptosi_sat(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value;