Auto merge of #76830 - Artoria2e5:tune, r=nagisa

Pass tune-cpu to LLVM

I think this is how it should work...

See https://internals.rust-lang.org/t/expose-tune-cpu-from-llvm/13088 for the background. Or the documentation diff.

4 files changed, 37 insertions, 6 deletions

diff --git a/compiler/rustc_codegen_llvm/src/attributes.rs b/compiler/rustc_codegen_llvm/src/attributes.rs
index f02c30c3ee3..8d131fa294b 100644
--- a/compiler/rustc_codegen_llvm/src/attributes.rs
+++ b/compiler/rustc_codegen_llvm/src/attributes.rs
@@ -194,6 +194,18 @@ pub fn apply_target_cpu_attr(cx: &CodegenCx<'ll, '_>, llfn: &'ll Value) {
     );
 }
 
+pub fn apply_tune_cpu_attr(cx: &CodegenCx<'ll, '_>, llfn: &'ll Value) {
+    if let Some(tune) = llvm_util::tune_cpu(cx.tcx.sess) {
+        let tune_cpu = SmallCStr::new(tune);
+        llvm::AddFunctionAttrStringValue(
+            llfn,
+            llvm::AttributePlace::Function,
+            const_cstr!("tune-cpu"),
+            tune_cpu.as_c_str(),
+        );
+    }
+}
+
 /// Sets the `NonLazyBind` LLVM attribute on a given function,
 /// assuming the codegen options allow skipping the PLT.
 pub fn non_lazy_bind(sess: &Session, llfn: &'ll Value) {
@@ -303,6 +315,9 @@ pub fn from_fn_attrs(cx: &CodegenCx<'ll, 'tcx>, llfn: &'ll Value, instance: ty::
     // Without this, ThinLTO won't inline Rust functions into Clang generated
     // functions (because Clang annotates functions this way too).
     apply_target_cpu_attr(cx, llfn);
+    // tune-cpu is only conveyed through the attribute for our purpose.
+    // The target doesn't care; the subtarget reads our attribute.
+    apply_tune_cpu_attr(cx, llfn);
 
     let features = llvm_target_features(cx.tcx.sess)
         .map(|s| s.to_string())
diff --git a/compiler/rustc_codegen_llvm/src/context.rs b/compiler/rustc_codegen_llvm/src/context.rs
index 1696f35563d..d3c71ff501b 100644
--- a/compiler/rustc_codegen_llvm/src/context.rs
+++ b/compiler/rustc_codegen_llvm/src/context.rs
@@ -417,7 +417,8 @@ impl MiscMethods<'tcx> for CodegenCx<'ll, 'tcx> {
     }
 
     fn apply_target_cpu_attr(&self, llfn: &'ll Value) {
-        attributes::apply_target_cpu_attr(self, llfn)
+        attributes::apply_target_cpu_attr(self, llfn);
+        attributes::apply_tune_cpu_attr(self, llfn);
     }
 
     fn create_used_variable(&self) {
diff --git a/compiler/rustc_codegen_llvm/src/lib.rs b/compiler/rustc_codegen_llvm/src/lib.rs
index 1237b39b300..b5f83118d0c 100644
--- a/compiler/rustc_codegen_llvm/src/lib.rs
+++ b/compiler/rustc_codegen_llvm/src/lib.rs
@@ -116,6 +116,9 @@ impl ExtraBackendMethods for LlvmCodegenBackend {
     fn target_cpu<'b>(&self, sess: &'b Session) -> &'b str {
         llvm_util::target_cpu(sess)
     }
+    fn tune_cpu<'b>(&self, sess: &'b Session) -> Option<&'b str> {
+        llvm_util::tune_cpu(sess)
+    }
 }
 
 impl WriteBackendMethods for LlvmCodegenBackend {
diff --git a/compiler/rustc_codegen_llvm/src/llvm_util.rs b/compiler/rustc_codegen_llvm/src/llvm_util.rs
index d42020047fd..a4605f46309 100644
--- a/compiler/rustc_codegen_llvm/src/llvm_util.rs
+++ b/compiler/rustc_codegen_llvm/src/llvm_util.rs
@@ -202,11 +202,7 @@ pub(crate) fn print(req: PrintRequest, sess: &Session) {
     }
 }
 
-pub fn target_cpu(sess: &Session) -> &str {
-    let name = match sess.opts.cg.target_cpu {
-        Some(ref s) => &**s,
-        None => &*sess.target.target.options.cpu,
-    };
+fn handle_native(name: &str) -> &str {
     if name != "native" {
         return name;
     }
@@ -217,3 +213,19 @@ pub fn target_cpu(sess: &Session) -> &str {
         str::from_utf8(slice::from_raw_parts(ptr as *const u8, len)).unwrap()
     }
 }
+
+pub fn target_cpu(sess: &Session) -> &str {
+    let name = match sess.opts.cg.target_cpu {
+        Some(ref s) => &**s,
+        None => &*sess.target.target.options.cpu,
+    };
+
+    handle_native(name)
+}
+
+pub fn tune_cpu(sess: &Session) -> Option<&str> {
+    match sess.opts.debugging_opts.tune_cpu {
+        Some(ref s) => Some(handle_native(&**s)),
+        None => None,
+    }
+}