Auto merge of #93048 - matthiaskrgr:rollup-cz5ma34, r=matthiaskrgr

Rollup of 9 pull requests

Successful merges:

 - #90782 (Implement raw-dylib support for windows-gnu)
 - #91150 (Let qpath contain NtTy: `<$:ty as $:ty>::…`)
 - #92425 (Improve SIMD casts)
 - #92692 (Simplify and unify rustdoc sidebar styles)
 - #92780 (Directly use ConstValue for single literals in blocks)
 - #92924 (Delete pretty printer tracing)
 - #93018 (Remove some unused `Ord` derives based on `Span`)
 - #93026 (fix typo in `max` description for f32/f64)
 - #93035 (Fix stdarch submodule pointing to commit outside tree)

Failed merges:

 - #92861 (Rustdoc mobile: put out-of-band info on its own line)

r? `@ghost`
`@rustbot` modify labels: rollup

18 files changed, 459 insertions, 311 deletions

diff --git a/compiler/rustc_ast/src/ast.rs b/compiler/rustc_ast/src/ast.rs
index d0732b35b6e..565488ab6a5 100644
--- a/compiler/rustc_ast/src/ast.rs
+++ b/compiler/rustc_ast/src/ast.rs
@@ -2006,7 +2006,7 @@ bitflags::bitflags! {
     }
 }
 
-#[derive(Clone, PartialEq, PartialOrd, Encodable, Decodable, Debug, Hash, HashStable_Generic)]
+#[derive(Clone, PartialEq, Encodable, Decodable, Debug, Hash, HashStable_Generic)]
 pub enum InlineAsmTemplatePiece {
     String(String),
     Placeholder { operand_idx: usize, modifier: Option<char>, span: Span },
@@ -2211,7 +2211,7 @@ pub enum IsAuto {
     No,
 }
 
-#[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable, Debug)]
+#[derive(Copy, Clone, PartialEq, Eq, Hash, Encodable, Decodable, Debug)]
 #[derive(HashStable_Generic)]
 pub enum Unsafe {
     Yes(Span),
diff --git a/compiler/rustc_ast_pretty/Cargo.toml b/compiler/rustc_ast_pretty/Cargo.toml
index 29f2be4cf46..5ad8714e9fe 100644
--- a/compiler/rustc_ast_pretty/Cargo.toml
+++ b/compiler/rustc_ast_pretty/Cargo.toml
@@ -7,6 +7,5 @@ edition = "2021"
 doctest = false
 
 [dependencies]
-tracing = "0.1"
 rustc_span = { path = "../rustc_span" }
 rustc_ast = { path = "../rustc_ast" }
diff --git a/compiler/rustc_ast_pretty/src/pp.rs b/compiler/rustc_ast_pretty/src/pp.rs
index ad9d15f1ce3..25437f8b53a 100644
--- a/compiler/rustc_ast_pretty/src/pp.rs
+++ b/compiler/rustc_ast_pretty/src/pp.rs
@@ -138,7 +138,6 @@ use ring::RingBuffer;
 use std::borrow::Cow;
 use std::collections::VecDeque;
 use std::fmt;
-use tracing::debug;
 
 /// How to break. Described in more detail in the module docs.
 #[derive(Clone, Copy, PartialEq)]
@@ -193,22 +192,6 @@ impl fmt::Display for Token {
     }
 }
 
-fn buf_str(buf: &RingBuffer<BufEntry>, left: usize, right: usize, lim: usize) -> String {
-    let mut i = left;
-    let mut l = lim;
-    let mut s = String::from("[");
-    while i != right && l != 0 {
-        l -= 1;
-        if i != left {
-            s.push_str(", ");
-        }
-        s.push_str(&format!("{}={}", buf[i].size, &buf[i].token));
-        i += 1;
-    }
-    s.push(']');
-    s
-}
-
 #[derive(Copy, Clone)]
 enum PrintStackBreak {
     Fits,
@@ -267,7 +250,6 @@ impl Default for BufEntry {
 impl Printer {
     pub fn new() -> Self {
         let linewidth = 78;
-        debug!("Printer::new {}", linewidth);
         let mut buf = RingBuffer::new();
         buf.advance_right();
         Printer {
@@ -310,16 +292,13 @@ impl Printer {
         } else {
             self.advance_right();
         }
-        debug!("pp Begin({})/buffer Vec<{},{}>", b.offset, self.left, self.right);
         self.scan_push(BufEntry { token: Token::Begin(b), size: -self.right_total });
     }
 
     fn scan_end(&mut self) {
         if self.scan_stack.is_empty() {
-            debug!("pp End/print Vec<{},{}>", self.left, self.right);
             self.print_end();
         } else {
-            debug!("pp End/buffer Vec<{},{}>", self.left, self.right);
             self.advance_right();
             self.scan_push(BufEntry { token: Token::End, size: -1 });
         }
@@ -334,7 +313,6 @@ impl Printer {
         } else {
             self.advance_right();
         }
-        debug!("pp Break({})/buffer Vec<{},{}>", b.offset, self.left, self.right);
         self.check_stack(0);
         self.scan_push(BufEntry { token: Token::Break(b), size: -self.right_total });
         self.right_total += b.blank_space;
@@ -342,10 +320,8 @@ impl Printer {
 
     fn scan_string(&mut self, s: Cow<'static, str>) {
         if self.scan_stack.is_empty() {
-            debug!("pp String('{}')/print Vec<{},{}>", s, self.left, self.right);
             self.print_string(s);
         } else {
-            debug!("pp String('{}')/buffer Vec<{},{}>", s, self.left, self.right);
             self.advance_right();
             let len = s.len() as isize;
             self.buf[self.right] = BufEntry { token: Token::String(s), size: len };
@@ -355,18 +331,8 @@ impl Printer {
     }
 
     fn check_stream(&mut self) {
-        debug!(
-            "check_stream Vec<{}, {}> with left_total={}, right_total={}",
-            self.left, self.right, self.left_total, self.right_total
-        );
         if self.right_total - self.left_total > self.space {
-            debug!(
-                "scan window is {}, longer than space on line ({})",
-                self.right_total - self.left_total,
-                self.space
-            );
             if Some(&self.left) == self.scan_stack.back() {
-                debug!("setting {} to infinity and popping", self.left);
                 let scanned = self.scan_pop_bottom();
                 self.buf[scanned].size = SIZE_INFINITY;
             }
@@ -378,7 +344,6 @@ impl Printer {
     }
 
     fn scan_push(&mut self, entry: BufEntry) {
-        debug!("scan_push {}", self.right);
         self.buf[self.right] = entry;
         self.scan_stack.push_front(self.right);
     }
@@ -401,11 +366,6 @@ impl Printer {
     }
 
     fn advance_left(&mut self) {
-        debug!(
-            "advance_left Vec<{},{}>, sizeof({})={}",
-            self.left, self.right, self.left, self.buf[self.left].size
-        );
-
         let mut left_size = self.buf[self.left].size;
 
         while left_size >= 0 {
@@ -465,14 +425,12 @@ impl Printer {
     }
 
     fn print_newline(&mut self, amount: isize) {
-        debug!("NEWLINE {}", amount);
         self.out.push('\n');
         self.pending_indentation = 0;
         self.indent(amount);
     }
 
     fn indent(&mut self, amount: isize) {
-        debug!("INDENT {}", amount);
         self.pending_indentation += amount;
     }
 
@@ -485,17 +443,14 @@ impl Printer {
     fn print_begin(&mut self, b: BeginToken, l: isize) {
         if l > self.space {
             let col = self.margin - self.space + b.offset;
-            debug!("print Begin -> push broken block at col {}", col);
             self.print_stack
                 .push(PrintStackElem { offset: col, pbreak: PrintStackBreak::Broken(b.breaks) });
         } else {
-            debug!("print Begin -> push fitting block");
             self.print_stack.push(PrintStackElem { offset: 0, pbreak: PrintStackBreak::Fits });
         }
     }
 
     fn print_end(&mut self) {
-        debug!("print End -> pop End");
         self.print_stack.pop().unwrap();
     }
 
@@ -503,22 +458,18 @@ impl Printer {
         let top = self.get_top();
         match top.pbreak {
             PrintStackBreak::Fits => {
-                debug!("print Break({}) in fitting block", b.blank_space);
                 self.space -= b.blank_space;
                 self.indent(b.blank_space);
             }
             PrintStackBreak::Broken(Breaks::Consistent) => {
-                debug!("print Break({}+{}) in consistent block", top.offset, b.offset);
                 self.print_newline(top.offset + b.offset);
                 self.space = self.margin - (top.offset + b.offset);
             }
             PrintStackBreak::Broken(Breaks::Inconsistent) => {
                 if l > self.space {
-                    debug!("print Break({}+{}) w/ newline in inconsistent", top.offset, b.offset);
                     self.print_newline(top.offset + b.offset);
                     self.space = self.margin - (top.offset + b.offset);
                 } else {
-                    debug!("print Break({}) w/o newline in inconsistent", b.blank_space);
                     self.indent(b.blank_space);
                     self.space -= b.blank_space;
                 }
@@ -528,7 +479,6 @@ impl Printer {
 
     fn print_string(&mut self, s: Cow<'static, str>) {
         let len = s.len() as isize;
-        debug!("print String({})", s);
         // assert!(len <= space);
         self.space -= len;
 
@@ -545,8 +495,6 @@ impl Printer {
     }
 
     fn print(&mut self, token: Token, l: isize) {
-        debug!("print {} {} (remaining line space={})", token, l, self.space);
-        debug!("{}", buf_str(&self.buf, self.left, self.right, 6));
         match token {
             Token::Begin(b) => self.print_begin(b, l),
             Token::End => self.print_end(),
diff --git a/compiler/rustc_codegen_llvm/src/back/archive.rs b/compiler/rustc_codegen_llvm/src/back/archive.rs
index 2fb5a0f9faf..5703a72c686 100644
--- a/compiler/rustc_codegen_llvm/src/back/archive.rs
+++ b/compiler/rustc_codegen_llvm/src/back/archive.rs
@@ -1,6 +1,7 @@
 //! A helper class for dealing with static archives
 
-use std::ffi::{CStr, CString};
+use std::env;
+use std::ffi::{CStr, CString, OsString};
 use std::io;
 use std::mem;
 use std::path::{Path, PathBuf};
@@ -158,54 +159,127 @@ impl<'a> ArchiveBuilder<'a> for LlvmArchiveBuilder<'a> {
             output_path.with_extension("lib")
         };
 
-        // we've checked for \0 characters in the library name already
-        let dll_name_z = CString::new(lib_name).unwrap();
-        // All import names are Rust identifiers and therefore cannot contain \0 characters.
-        // FIXME: when support for #[link_name] implemented, ensure that import.name values don't
-        // have any \0 characters
-        let import_name_and_ordinal_vector: Vec<(CString, Option<u16>)> = dll_imports
+        let mingw_gnu_toolchain = self.config.sess.target.llvm_target.ends_with("pc-windows-gnu");
+
+        let import_name_and_ordinal_vector: Vec<(String, Option<u16>)> = dll_imports
             .iter()
             .map(|import: &DllImport| {
                 if self.config.sess.target.arch == "x86" {
-                    (LlvmArchiveBuilder::i686_decorated_name(import), import.ordinal)
+                    (
+                        LlvmArchiveBuilder::i686_decorated_name(import, mingw_gnu_toolchain),
+                        import.ordinal,
+                    )
                 } else {
-                    (CString::new(import.name.to_string()).unwrap(), import.ordinal)
+                    (import.name.to_string(), import.ordinal)
                 }
             })
             .collect();
 
-        let output_path_z = rustc_fs_util::path_to_c_string(&output_path);
+        if mingw_gnu_toolchain {
+            // The binutils linker used on -windows-gnu targets cannot read the import
+            // libraries generated by LLVM: in our attempts, the linker produced an .EXE
+            // that loaded but crashed with an AV upon calling one of the imported
+            // functions.  Therefore, use binutils to create the import library instead,
+            // by writing a .DEF file to the temp dir and calling binutils's dlltool.
+            let def_file_path =
+                tmpdir.as_ref().join(format!("{}_imports", lib_name)).with_extension("def");
+
+            let def_file_content = format!(
+                "EXPORTS\n{}",
+                import_name_and_ordinal_vector
+                    .into_iter()
+                    .map(|(name, ordinal)| {
+                        match ordinal {
+                            Some(n) => format!("{} @{} NONAME", name, n),
+                            None => name,
+                        }
+                    })
+                    .collect::<Vec<String>>()
+                    .join("\n")
+            );
 
-        tracing::trace!("invoking LLVMRustWriteImportLibrary");
-        tracing::trace!("  dll_name {:#?}", dll_name_z);
-        tracing::trace!("  output_path {}", output_path.display());
-        tracing::trace!(
-            "  import names: {}",
-            dll_imports.iter().map(|import| import.name.to_string()).collect::<Vec<_>>().join(", "),
-        );
+            match std::fs::write(&def_file_path, def_file_content) {
+                Ok(_) => {}
+                Err(e) => {
+                    self.config.sess.fatal(&format!("Error writing .DEF file: {}", e));
+                }
+            };
 
-        let ffi_exports: Vec<LLVMRustCOFFShortExport> = import_name_and_ordinal_vector
-            .iter()
-            .map(|(name_z, ordinal)| LLVMRustCOFFShortExport::new(name_z.as_ptr(), *ordinal))
-            .collect();
-        let result = unsafe {
-            crate::llvm::LLVMRustWriteImportLibrary(
-                dll_name_z.as_ptr(),
-                output_path_z.as_ptr(),
-                ffi_exports.as_ptr(),
-                ffi_exports.len(),
-                llvm_machine_type(&self.config.sess.target.arch) as u16,
-                !self.config.sess.target.is_like_msvc,
-            )
-        };
+            let dlltool = find_binutils_dlltool(self.config.sess);
+            let result = std::process::Command::new(dlltool)
+                .args([
+                    "-d",
+                    def_file_path.to_str().unwrap(),
+                    "-D",
+                    lib_name,
+                    "-l",
+                    output_path.to_str().unwrap(),
+                ])
+                .output();
+
+            match result {
+                Err(e) => {
+                    self.config.sess.fatal(&format!("Error calling dlltool: {}", e.to_string()));
+                }
+                Ok(output) if !output.status.success() => self.config.sess.fatal(&format!(
+                    "Dlltool could not create import library: {}\n{}",
+                    String::from_utf8_lossy(&output.stdout),
+                    String::from_utf8_lossy(&output.stderr)
+                )),
+                _ => {}
+            }
+        } else {
+            // we've checked for \0 characters in the library name already
+            let dll_name_z = CString::new(lib_name).unwrap();
+
+            let output_path_z = rustc_fs_util::path_to_c_string(&output_path);
+
+            tracing::trace!("invoking LLVMRustWriteImportLibrary");
+            tracing::trace!("  dll_name {:#?}", dll_name_z);
+            tracing::trace!("  output_path {}", output_path.display());
+            tracing::trace!(
+                "  import names: {}",
+                dll_imports
+                    .iter()
+                    .map(|import| import.name.to_string())
+                    .collect::<Vec<_>>()
+                    .join(", "),
+            );
 
-        if result == crate::llvm::LLVMRustResult::Failure {
-            self.config.sess.fatal(&format!(
-                "Error creating import library for {}: {}",
-                lib_name,
-                llvm::last_error().unwrap_or("unknown LLVM error".to_string())
-            ));
-        }
+            // All import names are Rust identifiers and therefore cannot contain \0 characters.
+            // FIXME: when support for #[link_name] is implemented, ensure that the import names
+            // still don't contain any \0 characters.  Also need to check that the names don't
+            // contain substrings like " @" or "NONAME" that are keywords or otherwise reserved
+            // in definition files.
+            let cstring_import_name_and_ordinal_vector: Vec<(CString, Option<u16>)> =
+                import_name_and_ordinal_vector
+                    .into_iter()
+                    .map(|(name, ordinal)| (CString::new(name).unwrap(), ordinal))
+                    .collect();
+
+            let ffi_exports: Vec<LLVMRustCOFFShortExport> = cstring_import_name_and_ordinal_vector
+                .iter()
+                .map(|(name_z, ordinal)| LLVMRustCOFFShortExport::new(name_z.as_ptr(), *ordinal))
+                .collect();
+            let result = unsafe {
+                crate::llvm::LLVMRustWriteImportLibrary(
+                    dll_name_z.as_ptr(),
+                    output_path_z.as_ptr(),
+                    ffi_exports.as_ptr(),
+                    ffi_exports.len(),
+                    llvm_machine_type(&self.config.sess.target.arch) as u16,
+                    !self.config.sess.target.is_like_msvc,
+                )
+            };
+
+            if result == crate::llvm::LLVMRustResult::Failure {
+                self.config.sess.fatal(&format!(
+                    "Error creating import library for {}: {}",
+                    lib_name,
+                    llvm::last_error().unwrap_or("unknown LLVM error".to_string())
+                ));
+            }
+        };
 
         self.add_archive(&output_path, |_| false).unwrap_or_else(|e| {
             self.config.sess.fatal(&format!(
@@ -332,22 +406,61 @@ impl<'a> LlvmArchiveBuilder<'a> {
         }
     }
 
-    fn i686_decorated_name(import: &DllImport) -> CString {
+    fn i686_decorated_name(import: &DllImport, mingw: bool) -> String {
         let name = import.name;
-        // We verified during construction that `name` does not contain any NULL characters, so the
-        // conversion to CString is guaranteed to succeed.
-        CString::new(match import.calling_convention {
-            DllCallingConvention::C => format!("_{}", name),
-            DllCallingConvention::Stdcall(arg_list_size) => format!("_{}@{}", name, arg_list_size),
+        let prefix = if mingw { "" } else { "_" };
+
+        match import.calling_convention {
+            DllCallingConvention::C => format!("{}{}", prefix, name),
+            DllCallingConvention::Stdcall(arg_list_size) => {
+                format!("{}{}@{}", prefix, name, arg_list_size)
+            }
             DllCallingConvention::Fastcall(arg_list_size) => format!("@{}@{}", name, arg_list_size),
             DllCallingConvention::Vectorcall(arg_list_size) => {
                 format!("{}@@{}", name, arg_list_size)
             }
-        })
-        .unwrap()
+        }
     }
 }
 
 fn string_to_io_error(s: String) -> io::Error {
     io::Error::new(io::ErrorKind::Other, format!("bad archive: {}", s))
 }
+
+fn find_binutils_dlltool(sess: &Session) -> OsString {
+    assert!(sess.target.options.is_like_windows && !sess.target.options.is_like_msvc);
+    if let Some(dlltool_path) = &sess.opts.debugging_opts.dlltool {
+        return dlltool_path.clone().into_os_string();
+    }
+
+    let mut tool_name: OsString = if sess.host.arch != sess.target.arch {
+        // We are cross-compiling, so we need the tool with the prefix matching our target
+        if sess.target.arch == "x86" {
+            "i686-w64-mingw32-dlltool"
+        } else {
+            "x86_64-w64-mingw32-dlltool"
+        }
+    } else {
+        // We are not cross-compiling, so we just want `dlltool`
+        "dlltool"
+    }
+    .into();
+
+    if sess.host.options.is_like_windows {
+        // If we're compiling on Windows, add the .exe suffix
+        tool_name.push(".exe");
+    }
+
+    // NOTE: it's not clear how useful it is to explicitly search PATH.
+    for dir in env::split_paths(&env::var_os("PATH").unwrap_or_default()) {
+        let full_path = dir.join(&tool_name);
+        if full_path.is_file() {
+            return full_path.into_os_string();
+        }
+    }
+
+    // The user didn't specify the location of the dlltool binary, and we weren't able
+    // to find the appropriate one on the PATH.  Just return the name of the tool
+    // and let the invocation fail with a hopefully useful error message.
+    tool_name
+}
diff --git a/compiler/rustc_codegen_llvm/src/builder.rs b/compiler/rustc_codegen_llvm/src/builder.rs
index 5217fa2758f..8a9450c20dd 100644
--- a/compiler/rustc_codegen_llvm/src/builder.rs
+++ b/compiler/rustc_codegen_llvm/src/builder.rs
@@ -731,27 +731,11 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
     }
 
     fn fptoui_sat(&mut self, val: &'ll Value, dest_ty: &'ll Type) -> Option<&'ll Value> {
-        if !self.fptoint_sat_broken_in_llvm() {
-            let src_ty = self.cx.val_ty(val);
-            let float_width = self.cx.float_width(src_ty);
-            let int_width = self.cx.int_width(dest_ty);
-            let name = format!("llvm.fptoui.sat.i{}.f{}", int_width, float_width);
-            return Some(self.call_intrinsic(&name, &[val]));
-        }
-
-        None
+        self.fptoint_sat(false, val, dest_ty)
     }
 
     fn fptosi_sat(&mut self, val: &'ll Value, dest_ty: &'ll Type) -> Option<&'ll Value> {
-        if !self.fptoint_sat_broken_in_llvm() {
-            let src_ty = self.cx.val_ty(val);
-            let float_width = self.cx.float_width(src_ty);
-            let int_width = self.cx.int_width(dest_ty);
-            let name = format!("llvm.fptosi.sat.i{}.f{}", int_width, float_width);
-            return Some(self.call_intrinsic(&name, &[val]));
-        }
-
-        None
+        self.fptoint_sat(true, val, dest_ty)
     }
 
     fn fptoui(&mut self, val: &'ll Value, dest_ty: &'ll Type) -> &'ll Value {
@@ -1455,4 +1439,43 @@ impl<'a, 'll, 'tcx> Builder<'a, 'll, 'tcx> {
             _ => false,
         }
     }
+
+    fn fptoint_sat(
+        &mut self,
+        signed: bool,
+        val: &'ll Value,
+        dest_ty: &'ll Type,
+    ) -> Option<&'ll Value> {
+        if !self.fptoint_sat_broken_in_llvm() {
+            let src_ty = self.cx.val_ty(val);
+            let (float_ty, int_ty, vector_length) = if self.cx.type_kind(src_ty) == TypeKind::Vector
+            {
+                assert_eq!(self.cx.vector_length(src_ty), self.cx.vector_length(dest_ty));
+                (
+                    self.cx.element_type(src_ty),
+                    self.cx.element_type(dest_ty),
+                    Some(self.cx.vector_length(src_ty)),
+                )
+            } else {
+                (src_ty, dest_ty, None)
+            };
+            let float_width = self.cx.float_width(float_ty);
+            let int_width = self.cx.int_width(int_ty);
+
+            let instr = if signed { "fptosi" } else { "fptoui" };
+            let name = if let Some(vector_length) = vector_length {
+                format!(
+                    "llvm.{}.sat.v{}i{}.v{}f{}",
+                    instr, vector_length, int_width, vector_length, float_width
+                )
+            } else {
+                format!("llvm.{}.sat.i{}.f{}", instr, int_width, float_width)
+            };
+            let f =
+                self.declare_cfn(&name, llvm::UnnamedAddr::No, self.type_func(&[src_ty], dest_ty));
+            Some(self.call(self.type_func(&[src_ty], dest_ty), f, &[val], None))
+        } else {
+            None
+        }
+    }
 }
diff --git a/compiler/rustc_codegen_llvm/src/intrinsic.rs b/compiler/rustc_codegen_llvm/src/intrinsic.rs
index cebb6d13c4e..5adfa18035a 100644
--- a/compiler/rustc_codegen_llvm/src/intrinsic.rs
+++ b/compiler/rustc_codegen_llvm/src/intrinsic.rs
@@ -1688,7 +1688,7 @@ unsupported {} from `{}` with element `{}` of size `{}` to `{}`"#,
     bitwise_red!(simd_reduce_all: vector_reduce_and, true);
     bitwise_red!(simd_reduce_any: vector_reduce_or, true);
 
-    if name == sym::simd_cast {
+    if name == sym::simd_cast || name == sym::simd_as {
         require_simd!(ret_ty, "return");
         let (out_len, out_elem) = ret_ty.simd_size_and_type(bx.tcx());
         require!(
@@ -1714,14 +1714,26 @@ unsupported {} from `{}` with element `{}` of size `{}` to `{}`"#,
         let (in_style, in_width) = match in_elem.kind() {
             // vectors of pointer-sized integers should've been
             // disallowed before here, so this unwrap is safe.
-            ty::Int(i) => (Style::Int(true), i.bit_width().unwrap()),
-            ty::Uint(u) => (Style::Int(false), u.bit_width().unwrap()),
+            ty::Int(i) => (
+                Style::Int(true),
+                i.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
+            ),
+            ty::Uint(u) => (
+                Style::Int(false),
+                u.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
+            ),
             ty::Float(f) => (Style::Float, f.bit_width()),
             _ => (Style::Unsupported, 0),
         };
         let (out_style, out_width) = match out_elem.kind() {
-            ty::Int(i) => (Style::Int(true), i.bit_width().unwrap()),
-            ty::Uint(u) => (Style::Int(false), u.bit_width().unwrap()),
+            ty::Int(i) => (
+                Style::Int(true),
+                i.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
+            ),
+            ty::Uint(u) => (
+                Style::Int(false),
+                u.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
+            ),
             ty::Float(f) => (Style::Float, f.bit_width()),
             _ => (Style::Unsupported, 0),
         };
@@ -1748,10 +1760,10 @@ unsupported {} from `{}` with element `{}` of size `{}` to `{}`"#,
                 });
             }
             (Style::Float, Style::Int(out_is_signed)) => {
-                return Ok(if out_is_signed {
-                    bx.fptosi(args[0].immediate(), llret_ty)
-                } else {
-                    bx.fptoui(args[0].immediate(), llret_ty)
+                return Ok(match (out_is_signed, name == sym::simd_as) {
+                    (false, false) => bx.fptoui(args[0].immediate(), llret_ty),
+                    (true, false) => bx.fptosi(args[0].immediate(), llret_ty),
+                    (_, true) => bx.cast_float_to_int(out_is_signed, args[0].immediate(), llret_ty),
                 });
             }
             (Style::Float, Style::Float) => {
diff --git a/compiler/rustc_codegen_ssa/src/mir/rvalue.rs b/compiler/rustc_codegen_ssa/src/mir/rvalue.rs
index 679c4576701..68decce82ab 100644
--- a/compiler/rustc_codegen_ssa/src/mir/rvalue.rs
+++ b/compiler/rustc_codegen_ssa/src/mir/rvalue.rs
@@ -3,11 +3,10 @@ use super::place::PlaceRef;
 use super::{FunctionCx, LocalRef};
 
 use crate::base;
-use crate::common::{self, IntPredicate, RealPredicate};
+use crate::common::{self, IntPredicate};
 use crate::traits::*;
 use crate::MemFlags;
 
-use rustc_apfloat::{ieee, Float, Round, Status};
 use rustc_middle::mir;
 use rustc_middle::ty::cast::{CastTy, IntTy};
 use rustc_middle::ty::layout::{HasTyCtxt, LayoutOf};
@@ -368,10 +367,10 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
                                 bx.inttoptr(usize_llval, ll_t_out)
                             }
                             (CastTy::Float, CastTy::Int(IntTy::I)) => {
-                                cast_float_to_int(&mut bx, true, llval, ll_t_in, ll_t_out)
+                                bx.cast_float_to_int(true, llval, ll_t_out)
                             }
                             (CastTy::Float, CastTy::Int(_)) => {
-                                cast_float_to_int(&mut bx, false, llval, ll_t_in, ll_t_out)
+                                bx.cast_float_to_int(false, llval, ll_t_out)
                             }
                             _ => bug!("unsupported cast: {:?} to {:?}", operand.layout.ty, cast.ty),
                         };
@@ -768,146 +767,3 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
         // (*) this is only true if the type is suitable
     }
 }
-
-fn cast_float_to_int<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
-    bx: &mut Bx,
-    signed: bool,
-    x: Bx::Value,
-    float_ty: Bx::Type,
-    int_ty: Bx::Type,
-) -> Bx::Value {
-    if let Some(false) = bx.cx().sess().opts.debugging_opts.saturating_float_casts {
-        return if signed { bx.fptosi(x, int_ty) } else { bx.fptoui(x, int_ty) };
-    }
-
-    let try_sat_result = if signed { bx.fptosi_sat(x, int_ty) } else { bx.fptoui_sat(x, int_ty) };
-    if let Some(try_sat_result) = try_sat_result {
-        return try_sat_result;
-    }
-
-    let int_width = bx.cx().int_width(int_ty);
-    let float_width = bx.cx().float_width(float_ty);
-    // LLVM's fpto[su]i returns undef when the input x is infinite, NaN, or does not fit into the
-    // destination integer type after rounding towards zero. This `undef` value can cause UB in
-    // safe code (see issue #10184), so we implement a saturating conversion on top of it:
-    // Semantically, the mathematical value of the input is rounded towards zero to the next
-    // mathematical integer, and then the result is clamped into the range of the destination
-    // integer type. Positive and negative infinity are mapped to the maximum and minimum value of
-    // the destination integer type. NaN is mapped to 0.
-    //
-    // Define f_min and f_max as the largest and smallest (finite) floats that are exactly equal to
-    // a value representable in int_ty.
-    // They are exactly equal to int_ty::{MIN,MAX} if float_ty has enough significand bits.
-    // Otherwise, int_ty::MAX must be rounded towards zero, as it is one less than a power of two.
-    // int_ty::MIN, however, is either zero or a negative power of two and is thus exactly
-    // representable. Note that this only works if float_ty's exponent range is sufficiently large.
-    // f16 or 256 bit integers would break this property. Right now the smallest float type is f32
-    // with exponents ranging up to 127, which is barely enough for i128::MIN = -2^127.
-    // On the other hand, f_max works even if int_ty::MAX is greater than float_ty::MAX. Because
-    // we're rounding towards zero, we just get float_ty::MAX (which is always an integer).
-    // This already happens today with u128::MAX = 2^128 - 1 > f32::MAX.
-    let int_max = |signed: bool, int_width: u64| -> u128 {
-        let shift_amount = 128 - int_width;
-        if signed { i128::MAX as u128 >> shift_amount } else { u128::MAX >> shift_amount }
-    };
-    let int_min = |signed: bool, int_width: u64| -> i128 {
-        if signed { i128::MIN >> (128 - int_width) } else { 0 }
-    };
-
-    let compute_clamp_bounds_single = |signed: bool, int_width: u64| -> (u128, u128) {
-        let rounded_min = ieee::Single::from_i128_r(int_min(signed, int_width), Round::TowardZero);
-        assert_eq!(rounded_min.status, Status::OK);
-        let rounded_max = ieee::Single::from_u128_r(int_max(signed, int_width), Round::TowardZero);
-        assert!(rounded_max.value.is_finite());
-        (rounded_min.value.to_bits(), rounded_max.value.to_bits())
-    };
-    let compute_clamp_bounds_double = |signed: bool, int_width: u64| -> (u128, u128) {
-        let rounded_min = ieee::Double::from_i128_r(int_min(signed, int_width), Round::TowardZero);
-        assert_eq!(rounded_min.status, Status::OK);
-        let rounded_max = ieee::Double::from_u128_r(int_max(signed, int_width), Round::TowardZero);
-        assert!(rounded_max.value.is_finite());
-        (rounded_min.value.to_bits(), rounded_max.value.to_bits())
-    };
-
-    let mut float_bits_to_llval = |bits| {
-        let bits_llval = match float_width {
-            32 => bx.cx().const_u32(bits as u32),
-            64 => bx.cx().const_u64(bits as u64),
-            n => bug!("unsupported float width {}", n),
-        };
-        bx.bitcast(bits_llval, float_ty)
-    };
-    let (f_min, f_max) = match float_width {
-        32 => compute_clamp_bounds_single(signed, int_width),
-        64 => compute_clamp_bounds_double(signed, int_width),
-        n => bug!("unsupported float width {}", n),
-    };
-    let f_min = float_bits_to_llval(f_min);
-    let f_max = float_bits_to_llval(f_max);
-    // To implement saturation, we perform the following steps:
-    //
-    // 1. Cast x to an integer with fpto[su]i. This may result in undef.
-    // 2. Compare x to f_min and f_max, and use the comparison results to select:
-    //  a) int_ty::MIN if x < f_min or x is NaN
-    //  b) int_ty::MAX if x > f_max
-    //  c) the result of fpto[su]i otherwise
-    // 3. If x is NaN, return 0.0, otherwise return the result of step 2.
-    //
-    // This avoids resulting undef because values in range [f_min, f_max] by definition fit into the
-    // destination type. It creates an undef temporary, but *producing* undef is not UB. Our use of
-    // undef does not introduce any non-determinism either.
-    // More importantly, the above procedure correctly implements saturating conversion.
-    // Proof (sketch):
-    // If x is NaN, 0 is returned by definition.
-    // Otherwise, x is finite or infinite and thus can be compared with f_min and f_max.
-    // This yields three cases to consider:
-    // (1) if x in [f_min, f_max], the result of fpto[su]i is returned, which agrees with
-    //     saturating conversion for inputs in that range.
-    // (2) if x > f_max, then x is larger than int_ty::MAX. This holds even if f_max is rounded
-    //     (i.e., if f_max < int_ty::MAX) because in those cases, nextUp(f_max) is already larger
-    //     than int_ty::MAX. Because x is larger than int_ty::MAX, the return value of int_ty::MAX
-    //     is correct.
-    // (3) if x < f_min, then x is smaller than int_ty::MIN. As shown earlier, f_min exactly equals
-    //     int_ty::MIN and therefore the return value of int_ty::MIN is correct.
-    // QED.
-
-    let int_max = bx.cx().const_uint_big(int_ty, int_max(signed, int_width));
-    let int_min = bx.cx().const_uint_big(int_ty, int_min(signed, int_width) as u128);
-    let zero = bx.cx().const_uint(int_ty, 0);
-
-    // Step 1 ...
-    let fptosui_result = if signed { bx.fptosi(x, int_ty) } else { bx.fptoui(x, int_ty) };
-    let less_or_nan = bx.fcmp(RealPredicate::RealULT, x, f_min);
-    let greater = bx.fcmp(RealPredicate::RealOGT, x, f_max);
-
-    // Step 2: We use two comparisons and two selects, with %s1 being the
-    // result:
-    //     %less_or_nan = fcmp ult %x, %f_min
-    //     %greater = fcmp olt %x, %f_max
-    //     %s0 = select %less_or_nan, int_ty::MIN, %fptosi_result
-    //     %s1 = select %greater, int_ty::MAX, %s0
-    // Note that %less_or_nan uses an *unordered* comparison. This
-    // comparison is true if the operands are not comparable (i.e., if x is
-    // NaN). The unordered comparison ensures that s1 becomes int_ty::MIN if
-    // x is NaN.
-    //
-    // Performance note: Unordered comparison can be lowered to a "flipped"
-    // comparison and a negation, and the negation can be merged into the
-    // select. Therefore, it not necessarily any more expensive than an
-    // ordered ("normal") comparison. Whether these optimizations will be
-    // performed is ultimately up to the backend, but at least x86 does
-    // perform them.
-    let s0 = bx.select(less_or_nan, int_min, fptosui_result);
-    let s1 = bx.select(greater, int_max, s0);
-
-    // Step 3: NaN replacement.
-    // For unsigned types, the above step already yielded int_ty::MIN == 0 if x is NaN.
-    // Therefore we only need to execute this step for signed integer types.
-    if signed {
-        // LLVM has no isNaN predicate, so we use (x == x) instead
-        let cmp = bx.fcmp(RealPredicate::RealOEQ, x, x);
-        bx.select(cmp, s1, zero)
-    } else {
-        s1
-    }
-}
diff --git a/compiler/rustc_codegen_ssa/src/traits/builder.rs b/compiler/rustc_codegen_ssa/src/traits/builder.rs
index 48d88095855..5a06fb46105 100644
--- a/compiler/rustc_codegen_ssa/src/traits/builder.rs
+++ b/compiler/rustc_codegen_ssa/src/traits/builder.rs
@@ -1,18 +1,21 @@
 use super::abi::AbiBuilderMethods;
 use super::asm::AsmBuilderMethods;
+use super::consts::ConstMethods;
 use super::coverageinfo::CoverageInfoBuilderMethods;
 use super::debuginfo::DebugInfoBuilderMethods;
 use super::intrinsic::IntrinsicCallMethods;
-use super::type_::ArgAbiMethods;
+use super::misc::MiscMethods;
+use super::type_::{ArgAbiMethods, BaseTypeMethods};
 use super::{HasCodegen, StaticBuilderMethods};
 
 use crate::common::{
-    AtomicOrdering, AtomicRmwBinOp, IntPredicate, RealPredicate, SynchronizationScope,
+    AtomicOrdering, AtomicRmwBinOp, IntPredicate, RealPredicate, SynchronizationScope, TypeKind,
 };
 use crate::mir::operand::OperandRef;
 use crate::mir::place::PlaceRef;
 use crate::MemFlags;
 
+use rustc_apfloat::{ieee, Float, Round, Status};
 use rustc_middle::ty::layout::{HasParamEnv, TyAndLayout};
 use rustc_middle::ty::Ty;
 use rustc_span::Span;
@@ -202,6 +205,179 @@ pub trait BuilderMethods<'a, 'tcx>:
     fn intcast(&mut self, val: Self::Value, dest_ty: Self::Type, is_signed: bool) -> Self::Value;
     fn pointercast(&mut self, val: Self::Value, dest_ty: Self::Type) -> Self::Value;
 
+    fn cast_float_to_int(
+        &mut self,
+        signed: bool,
+        x: Self::Value,
+        dest_ty: Self::Type,
+    ) -> Self::Value {
+        let in_ty = self.cx().val_ty(x);
+        let (float_ty, int_ty) = if self.cx().type_kind(dest_ty) == TypeKind::Vector
+            && self.cx().type_kind(in_ty) == TypeKind::Vector
+        {
+            (self.cx().element_type(in_ty), self.cx().element_type(dest_ty))
+        } else {
+            (in_ty, dest_ty)
+        };
+        assert!(matches!(self.cx().type_kind(float_ty), TypeKind::Float | TypeKind::Double));
+        assert_eq!(self.cx().type_kind(int_ty), TypeKind::Integer);
+
+        if let Some(false) = self.cx().sess().opts.debugging_opts.saturating_float_casts {
+            return if signed { self.fptosi(x, dest_ty) } else { self.fptoui(x, dest_ty) };
+        }
+
+        let try_sat_result =
+            if signed { self.fptosi_sat(x, dest_ty) } else { self.fptoui_sat(x, dest_ty) };
+        if let Some(try_sat_result) = try_sat_result {
+            return try_sat_result;
+        }
+
+        let int_width = self.cx().int_width(int_ty);
+        let float_width = self.cx().float_width(float_ty);
+        // LLVM's fpto[su]i returns undef when the input x is infinite, NaN, or does not fit into the
+        // destination integer type after rounding towards zero. This `undef` value can cause UB in
+        // safe code (see issue #10184), so we implement a saturating conversion on top of it:
+        // Semantically, the mathematical value of the input is rounded towards zero to the next
+        // mathematical integer, and then the result is clamped into the range of the destination
+        // integer type. Positive and negative infinity are mapped to the maximum and minimum value of
+        // the destination integer type. NaN is mapped to 0.
+        //
+        // Define f_min and f_max as the largest and smallest (finite) floats that are exactly equal to
+        // a value representable in int_ty.
+        // They are exactly equal to int_ty::{MIN,MAX} if float_ty has enough significand bits.
+        // Otherwise, int_ty::MAX must be rounded towards zero, as it is one less than a power of two.
+        // int_ty::MIN, however, is either zero or a negative power of two and is thus exactly
+        // representable. Note that this only works if float_ty's exponent range is sufficiently large.
+        // f16 or 256 bit integers would break this property. Right now the smallest float type is f32
+        // with exponents ranging up to 127, which is barely enough for i128::MIN = -2^127.
+        // On the other hand, f_max works even if int_ty::MAX is greater than float_ty::MAX. Because
+        // we're rounding towards zero, we just get float_ty::MAX (which is always an integer).
+        // This already happens today with u128::MAX = 2^128 - 1 > f32::MAX.
+        let int_max = |signed: bool, int_width: u64| -> u128 {
+            let shift_amount = 128 - int_width;
+            if signed { i128::MAX as u128 >> shift_amount } else { u128::MAX >> shift_amount }
+        };
+        let int_min = |signed: bool, int_width: u64| -> i128 {
+            if signed { i128::MIN >> (128 - int_width) } else { 0 }
+        };
+
+        let compute_clamp_bounds_single = |signed: bool, int_width: u64| -> (u128, u128) {
+            let rounded_min =
+                ieee::Single::from_i128_r(int_min(signed, int_width), Round::TowardZero);
+            assert_eq!(rounded_min.status, Status::OK);
+            let rounded_max =
+                ieee::Single::from_u128_r(int_max(signed, int_width), Round::TowardZero);
+            assert!(rounded_max.value.is_finite());
+            (rounded_min.value.to_bits(), rounded_max.value.to_bits())
+        };
+        let compute_clamp_bounds_double = |signed: bool, int_width: u64| -> (u128, u128) {
+            let rounded_min =
+                ieee::Double::from_i128_r(int_min(signed, int_width), Round::TowardZero);
+            assert_eq!(rounded_min.status, Status::OK);
+            let rounded_max =
+                ieee::Double::from_u128_r(int_max(signed, int_width), Round::TowardZero);
+            assert!(rounded_max.value.is_finite());
+            (rounded_min.value.to_bits(), rounded_max.value.to_bits())
+        };
+        // To implement saturation, we perform the following steps:
+        //
+        // 1. Cast x to an integer with fpto[su]i. This may result in undef.
+        // 2. Compare x to f_min and f_max, and use the comparison results to select:
+        //  a) int_ty::MIN if x < f_min or x is NaN
+        //  b) int_ty::MAX if x > f_max
+        //  c) the result of fpto[su]i otherwise
+        // 3. If x is NaN, return 0.0, otherwise return the result of step 2.
+        //
+        // This avoids resulting undef because values in range [f_min, f_max] by definition fit into the
+        // destination type. It creates an undef temporary, but *producing* undef is not UB. Our use of
+        // undef does not introduce any non-determinism either.
+        // More importantly, the above procedure correctly implements saturating conversion.
+        // Proof (sketch):
+        // If x is NaN, 0 is returned by definition.
+        // Otherwise, x is finite or infinite and thus can be compared with f_min and f_max.
+        // This yields three cases to consider:
+        // (1) if x in [f_min, f_max], the result of fpto[su]i is returned, which agrees with
+        //     saturating conversion for inputs in that range.
+        // (2) if x > f_max, then x is larger than int_ty::MAX. This holds even if f_max is rounded
+        //     (i.e., if f_max < int_ty::MAX) because in those cases, nextUp(f_max) is already larger
+        //     than int_ty::MAX. Because x is larger than int_ty::MAX, the return value of int_ty::MAX
+        //     is correct.
+        // (3) if x < f_min, then x is smaller than int_ty::MIN. As shown earlier, f_min exactly equals
+        //     int_ty::MIN and therefore the return value of int_ty::MIN is correct.
+        // QED.
+
+        let float_bits_to_llval = |bx: &mut Self, bits| {
+            let bits_llval = match float_width {
+                32 => bx.cx().const_u32(bits as u32),
+                64 => bx.cx().const_u64(bits as u64),
+                n => bug!("unsupported float width {}", n),
+            };
+            bx.bitcast(bits_llval, float_ty)
+        };
+        let (f_min, f_max) = match float_width {
+            32 => compute_clamp_bounds_single(signed, int_width),
+            64 => compute_clamp_bounds_double(signed, int_width),
+            n => bug!("unsupported float width {}", n),
+        };
+        let f_min = float_bits_to_llval(self, f_min);
+        let f_max = float_bits_to_llval(self, f_max);
+        let int_max = self.cx().const_uint_big(int_ty, int_max(signed, int_width));
+        let int_min = self.cx().const_uint_big(int_ty, int_min(signed, int_width) as u128);
+        let zero = self.cx().const_uint(int_ty, 0);
+
+        // If we're working with vectors, constants must be "splatted": the constant is duplicated
+        // into each lane of the vector.  The algorithm stays the same, we are just using the
+        // same constant across all lanes.
+        let maybe_splat = |bx: &mut Self, val| {
+            if bx.cx().type_kind(dest_ty) == TypeKind::Vector {
+                bx.vector_splat(bx.vector_length(dest_ty), val)
+            } else {
+                val
+            }
+        };
+        let f_min = maybe_splat(self, f_min);
+        let f_max = maybe_splat(self, f_max);
+        let int_max = maybe_splat(self, int_max);
+        let int_min = maybe_splat(self, int_min);
+        let zero = maybe_splat(self, zero);
+
+        // Step 1 ...
+        let fptosui_result = if signed { self.fptosi(x, dest_ty) } else { self.fptoui(x, dest_ty) };
+        let less_or_nan = self.fcmp(RealPredicate::RealULT, x, f_min);
+        let greater = self.fcmp(RealPredicate::RealOGT, x, f_max);
+
+        // Step 2: We use two comparisons and two selects, with %s1 being the
+        // result:
+        //     %less_or_nan = fcmp ult %x, %f_min
+        //     %greater = fcmp olt %x, %f_max
+        //     %s0 = select %less_or_nan, int_ty::MIN, %fptosi_result
+        //     %s1 = select %greater, int_ty::MAX, %s0
+        // Note that %less_or_nan uses an *unordered* comparison. This
+        // comparison is true if the operands are not comparable (i.e., if x is
+        // NaN). The unordered comparison ensures that s1 becomes int_ty::MIN if
+        // x is NaN.
+        //
+        // Performance note: Unordered comparison can be lowered to a "flipped"
+        // comparison and a negation, and the negation can be merged into the
+        // select. Therefore, it not necessarily any more expensive than an
+        // ordered ("normal") comparison. Whether these optimizations will be
+        // performed is ultimately up to the backend, but at least x86 does
+        // perform them.
+        let s0 = self.select(less_or_nan, int_min, fptosui_result);
+        let s1 = self.select(greater, int_max, s0);
+
+        // Step 3: NaN replacement.
+        // For unsigned types, the above step already yielded int_ty::MIN == 0 if x is NaN.
+        // Therefore we only need to execute this step for signed integer types.
+        if signed {
+            // LLVM has no isNaN predicate, so we use (x == x) instead
+            let cmp = self.fcmp(RealPredicate::RealOEQ, x, x);
+            self.select(cmp, s1, zero)
+        } else {
+            s1
+        }
+    }
+
     fn icmp(&mut self, op: IntPredicate, lhs: Self::Value, rhs: Self::Value) -> Self::Value;
     fn fcmp(&mut self, op: RealPredicate, lhs: Self::Value, rhs: Self::Value) -> Self::Value;
 
diff --git a/compiler/rustc_interface/src/tests.rs b/compiler/rustc_interface/src/tests.rs
index 816e770f012..0e27a82b2b1 100644
--- a/compiler/rustc_interface/src/tests.rs
+++ b/compiler/rustc_interface/src/tests.rs
@@ -646,6 +646,7 @@ fn test_debugging_options_tracking_hash() {
     untracked!(borrowck, String::from("other"));
     untracked!(deduplicate_diagnostics, false);
     untracked!(dep_tasks, true);
+    untracked!(dlltool, Some(PathBuf::from("custom_dlltool.exe")));
     untracked!(dont_buffer_diagnostics, true);
     untracked!(dump_dep_graph, true);
     untracked!(dump_mir, Some(String::from("abc")));
diff --git a/compiler/rustc_metadata/src/native_libs.rs b/compiler/rustc_metadata/src/native_libs.rs
index 639d2e617c7..13cd8e4a046 100644
--- a/compiler/rustc_metadata/src/native_libs.rs
+++ b/compiler/rustc_metadata/src/native_libs.rs
@@ -274,11 +274,6 @@ impl Collector<'_> {
                     span,
                     "`#[link(...)]` with `kind = \"raw-dylib\"` only supported on Windows",
                 );
-            } else if !self.tcx.sess.target.options.is_like_msvc {
-                self.tcx.sess.span_warn(
-                    span,
-                    "`#[link(...)]` with `kind = \"raw-dylib\"` not supported on windows-gnu",
-                );
             }
 
             if lib_name.as_str().contains('\0') {
diff --git a/compiler/rustc_middle/src/mir/terminator.rs b/compiler/rustc_middle/src/mir/terminator.rs
index 51e4afaf220..fafd847a1cb 100644
--- a/compiler/rustc_middle/src/mir/terminator.rs
+++ b/compiler/rustc_middle/src/mir/terminator.rs
@@ -105,7 +105,7 @@ impl<'a> Iterator for SwitchTargetsIter<'a> {
 
 impl<'a> ExactSizeIterator for SwitchTargetsIter<'a> {}
 
-#[derive(Clone, TyEncodable, TyDecodable, Hash, HashStable, PartialEq, PartialOrd)]
+#[derive(Clone, TyEncodable, TyDecodable, Hash, HashStable, PartialEq)]
 pub enum TerminatorKind<'tcx> {
     /// Block should have one successor in the graph; we jump there.
     Goto { target: BasicBlock },
diff --git a/compiler/rustc_middle/src/ty/consts.rs b/compiler/rustc_middle/src/ty/consts.rs
index d6c35dfef88..19a73732fca 100644
--- a/compiler/rustc_middle/src/ty/consts.rs
+++ b/compiler/rustc_middle/src/ty/consts.rs
@@ -36,6 +36,7 @@ impl<'tcx> Const<'tcx> {
         Self::from_opt_const_arg_anon_const(tcx, ty::WithOptConstParam::unknown(def_id))
     }
 
+    #[instrument(skip(tcx), level = "debug")]
     pub fn from_opt_const_arg_anon_const(
         tcx: TyCtxt<'tcx>,
         def: ty::WithOptConstParam<LocalDefId>,
@@ -51,6 +52,7 @@ impl<'tcx> Const<'tcx> {
         };
 
         let expr = &tcx.hir().body(body_id).value;
+        debug!(?expr);
 
         let ty = tcx.type_of(def.def_id_for_type_of());
 
@@ -67,11 +69,21 @@ impl<'tcx> Const<'tcx> {
         }
     }
 
+    #[instrument(skip(tcx), level = "debug")]
     fn try_eval_lit_or_param(
         tcx: TyCtxt<'tcx>,
         ty: Ty<'tcx>,
         expr: &'tcx hir::Expr<'tcx>,
     ) -> Option<&'tcx Self> {
+        // Unwrap a block, so that e.g. `{ P }` is recognised as a parameter. Const arguments
+        // currently have to be wrapped in curly brackets, so it's necessary to special-case.
+        let expr = match &expr.kind {
+            hir::ExprKind::Block(block, _) if block.stmts.is_empty() && block.expr.is_some() => {
+                block.expr.as_ref().unwrap()
+            }
+            _ => expr,
+        };
+
         let lit_input = match expr.kind {
             hir::ExprKind::Lit(ref lit) => Some(LitToConstInput { lit: &lit.node, ty, neg: false }),
             hir::ExprKind::Unary(hir::UnOp::Neg, ref expr) => match expr.kind {
@@ -97,15 +109,6 @@ impl<'tcx> Const<'tcx> {
             }
         }
 
-        // Unwrap a block, so that e.g. `{ P }` is recognised as a parameter. Const arguments
-        // currently have to be wrapped in curly brackets, so it's necessary to special-case.
-        let expr = match &expr.kind {
-            hir::ExprKind::Block(block, _) if block.stmts.is_empty() && block.expr.is_some() => {
-                block.expr.as_ref().unwrap()
-            }
-            _ => expr,
-        };
-
         use hir::{def::DefKind::ConstParam, def::Res, ExprKind, Path, QPath};
         match expr.kind {
             ExprKind::Path(QPath::Resolved(_, &Path { res: Res::Def(ConstParam, def_id), .. })) => {
diff --git a/compiler/rustc_parse/src/parser/path.rs b/compiler/rustc_parse/src/parser/path.rs
index 4e60b7593c6..48502112e3a 100644
--- a/compiler/rustc_parse/src/parser/path.rs
+++ b/compiler/rustc_parse/src/parser/path.rs
@@ -139,22 +139,46 @@ impl<'a> Parser<'a> {
         style: PathStyle,
         ty_generics: Option<&Generics>,
     ) -> PResult<'a, Path> {
-        maybe_whole!(self, NtPath, |path| {
+        let reject_generics_if_mod_style = |parser: &Parser<'_>, path: &Path| {
+            // Ensure generic arguments don't end up in attribute paths, such as:
+            //
+            //     macro_rules! m {
+            //         ($p:path) => { #[$p] struct S; }
+            //     }
+            //
+            //     m!(inline<u8>); //~ ERROR: unexpected generic arguments in path
+            //
             if style == PathStyle::Mod && path.segments.iter().any(|segment| segment.args.is_some())
             {
-                self.struct_span_err(
-                    path.segments
-                        .iter()
-                        .filter_map(|segment| segment.args.as_ref())
-                        .map(|arg| arg.span())
-                        .collect::<Vec<_>>(),
-                    "unexpected generic arguments in path",
-                )
-                .emit();
+                parser
+                    .struct_span_err(
+                        path.segments
+                            .iter()
+                            .filter_map(|segment| segment.args.as_ref())
+                            .map(|arg| arg.span())
+                            .collect::<Vec<_>>(),
+                        "unexpected generic arguments in path",
+                    )
+                    .emit();
             }
+        };
+
+        maybe_whole!(self, NtPath, |path| {
+            reject_generics_if_mod_style(self, &path);
             path
         });
 
+        if let token::Interpolated(nt) = &self.token.kind {
+            if let token::NtTy(ty) = &**nt {
+                if let ast::TyKind::Path(None, path) = &ty.kind {
+                    let path = path.clone();
+                    self.bump();
+                    reject_generics_if_mod_style(self, &path);
+                    return Ok(path);
+                }
+            }
+        }
+
         let lo = self.token.span;
         let mut segments = Vec::new();
         let mod_sep_ctxt = self.token.span.ctxt();
diff --git a/compiler/rustc_session/src/options.rs b/compiler/rustc_session/src/options.rs
index 0b9623d1c7d..c48d8d689c1 100644
--- a/compiler/rustc_session/src/options.rs
+++ b/compiler/rustc_session/src/options.rs
@@ -1161,6 +1161,8 @@ options! {
     dep_tasks: bool = (false, parse_bool, [UNTRACKED],
         "print tasks that execute and the color their dep node gets (requires debug build) \
         (default: no)"),
+    dlltool: Option<PathBuf> = (None, parse_opt_pathbuf, [UNTRACKED],
+        "import library generation tool (windows-gnu only)"),
     dont_buffer_diagnostics: bool = (false, parse_bool, [UNTRACKED],
         "emit diagnostics rather than buffering (breaks NLL error downgrading, sorting) \
         (default: no)"),
diff --git a/compiler/rustc_span/src/symbol.rs b/compiler/rustc_span/src/symbol.rs
index af87399ac95..702e3594660 100644
--- a/compiler/rustc_span/src/symbol.rs
+++ b/compiler/rustc_span/src/symbol.rs
@@ -1223,6 +1223,7 @@ symbols! {
         simd,
         simd_add,
         simd_and,
+        simd_as,
         simd_bitmask,
         simd_cast,
         simd_ceil,
diff --git a/compiler/rustc_trait_selection/src/traits/coherence.rs b/compiler/rustc_trait_selection/src/traits/coherence.rs
index d174e00df77..af3540386f9 100644
--- a/compiler/rustc_trait_selection/src/traits/coherence.rs
+++ b/compiler/rustc_trait_selection/src/traits/coherence.rs
@@ -53,6 +53,7 @@ pub fn add_placeholder_note(err: &mut rustc_errors::DiagnosticBuilder<'_>) {
 /// If there are types that satisfy both impls, invokes `on_overlap`
 /// with a suitably-freshened `ImplHeader` with those types
 /// substituted. Otherwise, invokes `no_overlap`.
+#[instrument(skip(tcx, skip_leak_check, on_overlap, no_overlap), level = "debug")]
 pub fn overlapping_impls<F1, F2, R>(
     tcx: TyCtxt<'_>,
     impl1_def_id: DefId,
@@ -65,12 +66,6 @@ where
     F1: FnOnce(OverlapResult<'_>) -> R,
     F2: FnOnce() -> R,
 {
-    debug!(
-        "overlapping_impls(\
-           impl1_def_id={:?}, \
-           impl2_def_id={:?})",
-        impl1_def_id, impl2_def_id,
-    );
     // Before doing expensive operations like entering an inference context, do
     // a quick check via fast_reject to tell if the impl headers could possibly
     // unify.
@@ -85,6 +80,7 @@ where
     .any(|(ty1, ty2)| {
         let t1 = fast_reject::simplify_type(tcx, ty1, SimplifyParams::No, StripReferences::No);
         let t2 = fast_reject::simplify_type(tcx, ty2, SimplifyParams::No, StripReferences::No);
+
         if let (Some(t1), Some(t2)) = (t1, t2) {
             // Simplified successfully
             t1 != t2
diff --git a/compiler/rustc_trait_selection/src/traits/specialize/mod.rs b/compiler/rustc_trait_selection/src/traits/specialize/mod.rs
index ab732f510ff..cd2e0f18e0c 100644
--- a/compiler/rustc_trait_selection/src/traits/specialize/mod.rs
+++ b/compiler/rustc_trait_selection/src/traits/specialize/mod.rs
@@ -117,9 +117,8 @@ pub fn translate_substs<'a, 'tcx>(
 /// Specialization is determined by the sets of types to which the impls apply;
 /// `impl1` specializes `impl2` if it applies to a subset of the types `impl2` applies
 /// to.
+#[instrument(skip(tcx), level = "debug")]
 pub(super) fn specializes(tcx: TyCtxt<'_>, (impl1_def_id, impl2_def_id): (DefId, DefId)) -> bool {
-    debug!("specializes({:?}, {:?})", impl1_def_id, impl2_def_id);
-
     // The feature gate should prevent introducing new specializations, but not
     // taking advantage of upstream ones.
     let features = tcx.features();
diff --git a/compiler/rustc_typeck/src/check/intrinsic.rs b/compiler/rustc_typeck/src/check/intrinsic.rs
index 6314f2aba4e..4c612ed5be5 100644
--- a/compiler/rustc_typeck/src/check/intrinsic.rs
+++ b/compiler/rustc_typeck/src/check/intrinsic.rs
@@ -453,7 +453,7 @@ pub fn check_platform_intrinsic_type(tcx: TyCtxt<'_>, it: &hir::ForeignItem<'_>)
         sym::simd_scatter => (3, vec![param(0), param(1), param(2)], tcx.mk_unit()),
         sym::simd_insert => (2, vec![param(0), tcx.types.u32, param(1)], param(0)),
         sym::simd_extract => (2, vec![param(0), tcx.types.u32], param(1)),
-        sym::simd_cast => (2, vec![param(0)], param(1)),
+        sym::simd_cast | sym::simd_as => (2, vec![param(0)], param(1)),
         sym::simd_bitmask => (2, vec![param(0)], param(1)),
         sym::simd_select | sym::simd_select_bitmask => {
             (2, vec![param(0), param(1), param(1)], param(1))