From 7e9dc9c1fdfa7ecffb61c109d646080ed49c12e3 Mon Sep 17 00:00:00 2001
From: Dan Klishch <danilklishch@gmail.com>
Date: Fri, 5 Jul 2024 12:46:18 -0400
Subject: [PATCH] AK: Use bit_cast in
 SIMDExtras.h/AK::Detail::byte_reverse_impl

This necessitates marking bit_cast as ALWAYS_INLINE since emitting it as
a function call there will create an unnecessary potential SSE
registers -> plain registers/memory round-trip.
---
 AK/BitCast.h    |  2 +-
 AK/SIMDExtras.h | 11 +++++------
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/AK/BitCast.h b/AK/BitCast.h
index a1f6f17d39f..32752c738f3 100644
--- a/AK/BitCast.h
+++ b/AK/BitCast.h
@@ -11,7 +11,7 @@
 namespace AK {
 
 template<typename T, typename U>
-[[nodiscard]] constexpr inline T bit_cast(U const& a)
+[[nodiscard]] constexpr ALWAYS_INLINE T bit_cast(U const& a)
 {
 #if (__has_builtin(__builtin_bit_cast))
     return __builtin_bit_cast(T, a);
diff --git a/AK/SIMDExtras.h b/AK/SIMDExtras.h
index d8568cf6f9c..3147f2fb3a8 100644
--- a/AK/SIMDExtras.h
+++ b/AK/SIMDExtras.h
@@ -261,12 +261,11 @@ ALWAYS_INLINE static T byte_reverse_impl(T a, IndexSequence<Idx...>)
     //        Hence this giant conditional
     using BytesVector = Conditional<sizeof(T) == 2, u8x2, Conditional<sizeof(T) == 4, u8x4, Conditional<sizeof(T) == 8, u8x8, Conditional<sizeof(T) == 16, u8x16, Conditional<sizeof(T) == 32, u8x32, void>>>>>;
     static_assert(sizeof(BytesVector) == sizeof(T));
-    // Note: Using __builtin_bit_cast instead of bit_cast to avoid a psabi warning from bit_cast
-    auto tmp = __builtin_shufflevector(
-        __builtin_bit_cast(BytesVector, a),
-        __builtin_bit_cast(BytesVector, a),
-        N - 1 - Idx...);
-    return __builtin_bit_cast(T, tmp);
+    return bit_cast<T>(
+        __builtin_shufflevector(
+            bit_cast<BytesVector>(a),
+            bit_cast<BytesVector>(a),
+            N - 1 - Idx...));
 }
 
 template<SIMDVector T, size_t... Idx>