Auto merge of #147404 - JamieCunliffe:inline-always, r=jackh726

Fix issue with callsite inline attribute not being applied sometimes.

If the calling function had more target features enabled than the callee than the attribute wasn't being applied as the arguments for the check had been swapped round. Also includes target features that are part of the global set as the warning was checking those but when adding the attribute they were not checked.

Add a codegen-llvm test to check that the attribute is actually applied as previously only the warning was being checked.

Tracking issue: rust-lang/rust#145574

compiler/rustc_codegen_llvm/src/builder.rs

@@ -16,7 +16,7 @@ use rustc_codegen_ssa::mir::place::PlaceRef;
 use rustc_codegen_ssa::traits::*;
 use rustc_data_structures::small_c_str::SmallCStr;
 use rustc_hir::def_id::DefId;
-use rustc_middle::middle::codegen_fn_attrs::CodegenFnAttrs;
+use rustc_middle::middle::codegen_fn_attrs::{CodegenFnAttrs, TargetFeature, TargetFeatureKind};
 use rustc_middle::ty::layout::{
     FnAbiError, FnAbiOfHelpers, FnAbiRequest, HasTypingEnv, LayoutError, LayoutOfHelpers,
     TyAndLayout,
@@ -1405,14 +1405,18 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
             // Attributes on the function definition being called
             let fn_defn_attrs = self.cx.tcx.codegen_fn_attrs(instance.def_id());
             if let Some(fn_call_attrs) = fn_call_attrs
-                && !fn_call_attrs.target_features.is_empty()
                 // If there is an inline attribute and a target feature that matches
                 // we will add the attribute to the callsite otherwise we'll omit
                 // this and not add the attribute to prevent soundness issues.
                 && let Some(inlining_rule) = attributes::inline_attr(&self.cx, self.cx.tcx, instance)
                 && self.cx.tcx.is_target_feature_call_safe(
-                    &fn_call_attrs.target_features,
                     &fn_defn_attrs.target_features,
+                    &fn_call_attrs.target_features.iter().cloned().chain(
+                        self.cx.tcx.sess.target_features.iter().map(|feat| TargetFeature {
+                            name: *feat,
+                            kind: TargetFeatureKind::Implied,
+                        })
+                    ).collect::<Vec<_>>(),
                 )
             {
                 attributes::apply_to_callsite(

compiler/rustc_middle/src/ty/context.rs

@@ -2090,6 +2090,8 @@ impl<'tcx> TyCtxt<'tcx> {
         self.sess.dcx()
     }
 
+    /// Checks to see if the caller (`body_features`) has all the features required by the callee
+    /// (`callee_features`).
     pub fn is_target_feature_call_safe(
         self,
         callee_features: &[TargetFeature],

tests/codegen-llvm/inline-always-callsite.rs

@@ -0,0 +1,41 @@
+//@ add-minicore
+//@ compile-flags: --target aarch64-unknown-linux-gnu -Zinline-mir=no -C no-prepopulate-passes
+//@ needs-llvm-components: aarch64
+
+#![crate_type = "lib"]
+#![feature(no_core, lang_items, target_feature_inline_always)]
+#![no_core]
+
+extern crate minicore;
+use minicore::*;
+
+#[inline(always)]
+#[target_feature(enable = "neon")]
+#[no_mangle]
+pub fn single_target_feature() -> i32 {
+    42
+}
+
+#[inline(always)]
+#[target_feature(enable = "neon,i8mm")]
+#[no_mangle]
+// CHECK: define{{( noundef)?}} i32 @multiple_target_features() unnamed_addr #1 {
+pub fn multiple_target_features() -> i32 {
+    // CHECK: %_0 = call{{( noundef)?}} i32 @single_target_feature() #3
+    single_target_feature()
+}
+
+#[no_mangle]
+// CHECK: define{{( noundef)?}} i32 @inherits_from_global() unnamed_addr #2 {
+pub fn inherits_from_global() -> i32 {
+    unsafe {
+        // CHECK: %_0 = call{{( noundef)?}} i32 @single_target_feature() #3
+        single_target_feature()
+    }
+}
+
+// Attribute #3 requires the alwaysinline attribute, the alwaysinline attribute is not emitted on a
+// function definition when target features are present, rather it will be moved onto the function
+// call, if the features match up.
+//
+// CHECK: attributes #3 = { alwaysinline nounwind }