From 807f77fb851c8323e3a3a80825514fd5c6a93cfb Mon Sep 17 00:00:00 2001
From: Artyom Pavlov <newpavlov@gmail.com>
Date: Mon, 12 Aug 2024 21:54:06 +0300
Subject: [PATCH] cpufeatures: use `#[cold]` for initialization code (#1096)

---
 Cargo.lock             | 21 ++++++++++++++-------
 cpufeatures/Cargo.toml |  2 +-
 cpufeatures/src/lib.rs | 21 ++++++++++++---------
 3 files changed, 27 insertions(+), 17 deletions(-)

diff --git a/Cargo.lock b/Cargo.lock
index 35562a76..308907a5 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -35,7 +35,7 @@ version = "0.0.2"
 
 [[package]]
 name = "cpufeatures"
-version = "0.2.12"
+version = "0.2.13"
 dependencies = [
  "libc",
 ]
@@ -111,6 +111,12 @@ version = "0.2.155"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "97b3888a4aecf77e811145cadf6eef5901f4782c53886191b2f693f24761847c"
 
+[[package]]
+name = "memchr"
+version = "2.7.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "78ca9ab1a0babb1e7d5695e3530886289c18cf2f87ec19a575a0abdce112e3a3"
+
 [[package]]
 name = "opaque-debug"
 version = "0.3.1"
@@ -151,18 +157,18 @@ checksum = "f3cb5ba0dc43242ce17de99c180e96db90b235b8a9fdc9543c96d2209116bd9f"
 
 [[package]]
 name = "serde"
-version = "1.0.203"
+version = "1.0.206"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7253ab4de971e72fb7be983802300c30b5a7f0c2e56fab8abfc6a214307c0094"
+checksum = "5b3e4cd94123dd520a128bcd11e34d9e9e423e7e3e50425cb1b4b1e3549d0284"
 dependencies = [
  "serde_derive",
 ]
 
 [[package]]
 name = "serde_derive"
-version = "1.0.203"
+version = "1.0.206"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "500cbc0ebeb6f46627f50f3f5811ccf6bf00643be300b4c3eabc0ef55dc5b5ba"
+checksum = "fabfb6138d2383ea8208cf98ccf69cdfb1aff4088460681d84189aa259762f97"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -171,11 +177,12 @@ dependencies = [
 
 [[package]]
 name = "serde_json"
-version = "1.0.117"
+version = "1.0.124"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "455182ea6142b14f93f4bc5320a2b31c1f266b66a4a5c858b013302a5d8cbfc3"
+checksum = "66ad62847a56b3dba58cc891acd13884b9c61138d330c0d7b6181713d4fce38d"
 dependencies = [
  "itoa",
+ "memchr",
  "ryu",
  "serde",
 ]
diff --git a/cpufeatures/Cargo.toml b/cpufeatures/Cargo.toml
index ff2c2329..a8af9ffa 100644
--- a/cpufeatures/Cargo.toml
+++ b/cpufeatures/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "cpufeatures"
-version = "0.2.12"
+version = "0.2.13"
 description = """
 Lightweight runtime CPU feature detection for aarch64, loongarch64, and x86/x86_64 targets, 
 with no_std support and support for mobile targets including Android and iOS
diff --git a/cpufeatures/src/lib.rs b/cpufeatures/src/lib.rs
index ba9aa25d..bc4afd9a 100644
--- a/cpufeatures/src/lib.rs
+++ b/cpufeatures/src/lib.rs
@@ -177,19 +177,24 @@ macro_rules! new {
                 }
             }
 
-            /// Initialize underlying storage if needed and get
-            /// stored value and initialization token.
+            /// Get stored value and initialization token,
+            /// initializing underlying storage if needed.
             #[inline]
             pub fn init_get() -> (InitToken, bool) {
                 let res = $crate::__unless_target_features! {
                     $($tf),+ => {
+                        #[cold]
+                        fn init_inner() -> bool {
+                            let res = $crate::__detect_target_features!($($tf),+);
+                            STORAGE.store(res as u8, Relaxed);
+                            res
+                        }
+
                         // Relaxed ordering is fine, as we only have a single atomic variable.
                         let val = STORAGE.load(Relaxed);
 
                         if val == UNINIT {
-                            let res = $crate::__detect_target_features!($($tf),+);
-                            STORAGE.store(res as u8, Relaxed);
-                            res
+                            init_inner()
                         } else {
                             val == 1
                         }
@@ -199,15 +204,13 @@ macro_rules! new {
                 (InitToken(()), res)
             }
 
-            /// Initialize underlying storage if needed and get
-            /// initialization token.
+            /// Initialize underlying storage if needed and get initialization token.
             #[inline]
             pub fn init() -> InitToken {
                 init_get().0
             }
 
-            /// Initialize underlying storage if needed and get
-            /// stored value.
+            /// Initialize underlying storage if needed and get stored value.
             #[inline]
             pub fn get() -> bool {
                 init_get().1