Auto merge of #17995 - lnicola:sync-from-rust, r=lnicola

minor: sync from downstream

5 files changed, 51 insertions, 34 deletions

diff --git a/compiler/rustc_codegen_gcc/example/mini_core_hello_world.rs b/compiler/rustc_codegen_gcc/example/mini_core_hello_world.rs
index 9f096e90220..dcfa34cb729 100644
--- a/compiler/rustc_codegen_gcc/example/mini_core_hello_world.rs
+++ b/compiler/rustc_codegen_gcc/example/mini_core_hello_world.rs
@@ -2,7 +2,7 @@
 
 #![feature(
     no_core, unboxed_closures, start, lang_items, never_type, linkage,
-    extern_types, thread_local, raw_ref_op
+    extern_types, thread_local
 )]
 #![no_core]
 #![allow(dead_code, internal_features, non_camel_case_types)]
diff --git a/compiler/rustc_codegen_gcc/src/builder.rs b/compiler/rustc_codegen_gcc/src/builder.rs
index 47b378cc1cd..6ba678e2e7c 100644
--- a/compiler/rustc_codegen_gcc/src/builder.rs
+++ b/compiler/rustc_codegen_gcc/src/builder.rs
@@ -1923,15 +1923,11 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
         v2: RValue<'gcc>,
         mask: RValue<'gcc>,
     ) -> RValue<'gcc> {
-        let struct_type = mask.get_type().is_struct().expect("mask should be of struct type");
-
         // TODO(antoyo): use a recursive unqualified() here.
         let vector_type = v1.get_type().unqualified().dyncast_vector().expect("vector type");
         let element_type = vector_type.get_element_type();
         let vec_num_units = vector_type.get_num_units();
 
-        let mask_num_units = struct_type.get_field_count();
-        let mut vector_elements = vec![];
         let mask_element_type = if element_type.is_integral() {
             element_type
         } else {
@@ -1942,19 +1938,39 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
             #[cfg(not(feature = "master"))]
             self.int_type
         };
-        for i in 0..mask_num_units {
-            let field = struct_type.get_field(i as i32);
-            vector_elements.push(self.context.new_cast(
-                self.location,
-                mask.access_field(self.location, field).to_rvalue(),
-                mask_element_type,
-            ));
-        }
+
+        let mut mask_elements = if let Some(vector_type) = mask.get_type().dyncast_vector() {
+            let mask_num_units = vector_type.get_num_units();
+            let mut mask_elements = vec![];
+            for i in 0..mask_num_units {
+                let index = self.context.new_rvalue_from_long(self.cx.type_u32(), i as _);
+                mask_elements.push(self.context.new_cast(
+                    self.location,
+                    self.extract_element(mask, index).to_rvalue(),
+                    mask_element_type,
+                ));
+            }
+            mask_elements
+        } else {
+            let struct_type = mask.get_type().is_struct().expect("mask should be of struct type");
+            let mask_num_units = struct_type.get_field_count();
+            let mut mask_elements = vec![];
+            for i in 0..mask_num_units {
+                let field = struct_type.get_field(i as i32);
+                mask_elements.push(self.context.new_cast(
+                    self.location,
+                    mask.access_field(self.location, field).to_rvalue(),
+                    mask_element_type,
+                ));
+            }
+            mask_elements
+        };
+        let mask_num_units = mask_elements.len();
 
         // NOTE: the mask needs to be the same length as the input vectors, so add the missing
         // elements in the mask if needed.
         for _ in mask_num_units..vec_num_units {
-            vector_elements.push(self.context.new_rvalue_zero(mask_element_type));
+            mask_elements.push(self.context.new_rvalue_zero(mask_element_type));
         }
 
         let result_type = self.context.new_vector_type(element_type, mask_num_units as u64);
@@ -1998,7 +2014,7 @@ impl<'a, 'gcc, 'tcx> Builder<'a, 'gcc, 'tcx> {
 
         let new_mask_num_units = std::cmp::max(mask_num_units, vec_num_units);
         let mask_type = self.context.new_vector_type(mask_element_type, new_mask_num_units as u64);
-        let mask = self.context.new_rvalue_from_vector(self.location, mask_type, &vector_elements);
+        let mask = self.context.new_rvalue_from_vector(self.location, mask_type, &mask_elements);
         let result = self.context.new_rvalue_vector_perm(self.location, v1, v2, mask);
 
         if vec_num_units != mask_num_units {
diff --git a/compiler/rustc_codegen_gcc/src/intrinsic/simd.rs b/compiler/rustc_codegen_gcc/src/intrinsic/simd.rs
index 8da1df3be15..96a833ccaf2 100644
--- a/compiler/rustc_codegen_gcc/src/intrinsic/simd.rs
+++ b/compiler/rustc_codegen_gcc/src/intrinsic/simd.rs
@@ -353,19 +353,24 @@ pub fn generic_simd_intrinsic<'a, 'gcc, 'tcx>(
     }
 
     if name == sym::simd_shuffle {
-        // Make sure this is actually an array, since typeck only checks the length-suffixed
+        // Make sure this is actually an array or SIMD vector, since typeck only checks the length-suffixed
         // version of this intrinsic.
-        let n: u64 = match *args[2].layout.ty.kind() {
+        let idx_ty = args[2].layout.ty;
+        let n: u64 = match idx_ty.kind() {
             ty::Array(ty, len) if matches!(*ty.kind(), ty::Uint(ty::UintTy::U32)) => {
                 len.try_eval_target_usize(bx.cx.tcx, ty::ParamEnv::reveal_all()).unwrap_or_else(
                     || span_bug!(span, "could not evaluate shuffle index array length"),
                 )
             }
-            _ => return_error!(InvalidMonomorphization::SimdShuffle {
-                span,
-                name,
-                ty: args[2].layout.ty
-            }),
+            _ if idx_ty.is_simd()
+                && matches!(
+                    idx_ty.simd_size_and_type(bx.cx.tcx).1.kind(),
+                    ty::Uint(ty::UintTy::U32)
+                ) =>
+            {
+                idx_ty.simd_size_and_type(bx.cx.tcx).0
+            }
+            _ => return_error!(InvalidMonomorphization::SimdShuffle { span, name, ty: idx_ty }),
         };
         require_simd!(ret_ty, InvalidMonomorphization::SimdReturn { span, name, ty: ret_ty });
 
diff --git a/compiler/rustc_codegen_gcc/src/type_of.rs b/compiler/rustc_codegen_gcc/src/type_of.rs
index d85ed4f12ff..b7b1be5369c 100644
--- a/compiler/rustc_codegen_gcc/src/type_of.rs
+++ b/compiler/rustc_codegen_gcc/src/type_of.rs
@@ -213,9 +213,8 @@ impl<'tcx> LayoutGccExt<'tcx> for TyAndLayout<'tcx> {
                 // NOTE: we cannot remove this match like in the LLVM codegen because the call
                 // to fn_ptr_backend_type handle the on-stack attribute.
                 // TODO(antoyo): find a less hackish way to hande the on-stack attribute.
-                ty::FnPtr(sig) => {
-                    cx.fn_ptr_backend_type(cx.fn_abi_of_fn_ptr(sig, ty::List::empty()))
-                }
+                ty::FnPtr(sig_tys, hdr) => cx
+                    .fn_ptr_backend_type(cx.fn_abi_of_fn_ptr(sig_tys.with(hdr), ty::List::empty())),
                 _ => self.scalar_gcc_type_at(cx, scalar, Size::ZERO),
             };
             cx.scalar_types.borrow_mut().insert(self.ty, ty);
diff --git a/compiler/rustc_codegen_gcc/tests/run/asm.rs b/compiler/rustc_codegen_gcc/tests/run/asm.rs
index 56f2aac3d0a..4e05d026868 100644
--- a/compiler/rustc_codegen_gcc/tests/run/asm.rs
+++ b/compiler/rustc_codegen_gcc/tests/run/asm.rs
@@ -3,12 +3,10 @@
 // Run-time:
 //   status: 0
 
-#![feature(asm_const)]
-
-#[cfg(target_arch="x86_64")]
+#[cfg(target_arch = "x86_64")]
 use std::arch::{asm, global_asm};
 
-#[cfg(target_arch="x86_64")]
+#[cfg(target_arch = "x86_64")]
 global_asm!(
     "
     .global add_asm
@@ -22,7 +20,7 @@ extern "C" {
     fn add_asm(a: i64, b: i64) -> i64;
 }
 
-#[cfg(target_arch="x86_64")]
+#[cfg(target_arch = "x86_64")]
 pub unsafe fn mem_cpy(dst: *mut u8, src: *const u8, len: usize) {
     asm!(
         "rep movsb",
@@ -33,7 +31,7 @@ pub unsafe fn mem_cpy(dst: *mut u8, src: *const u8, len: usize) {
     );
 }
 
-#[cfg(target_arch="x86_64")]
+#[cfg(target_arch = "x86_64")]
 fn asm() {
     unsafe {
         asm!("nop");
@@ -178,9 +176,8 @@ fn asm() {
     assert_eq!(array1, array2);
 }
 
-#[cfg(not(target_arch="x86_64"))]
-fn asm() {
-}
+#[cfg(not(target_arch = "x86_64"))]
+fn asm() {}
 
 fn main() {
     asm();