From: Nathan Moin Vaziri <nathan@nathanm.com>
Date: Fri, 17 Apr 2026 20:23:52 +0000 (-0700)
Subject: Merge duplicate 32-bit _mm_cvtsi64_si128 polyfills
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=4c5cc72e496833f2b2a7db495ca19a5b326924a6;p=thirdparty%2Fzlib-ng.git

Merge duplicate 32-bit _mm_cvtsi64_si128 polyfills

The MSVC and GCC 32-bit polyfills for _mm_cvtsi64_si128 /
_mm_cvtsi128_si64 had identical bodies. Merge them into a single
block guarded by !__clang__ && ARCH_32BIT, with the MSVC-only
#include <intrin.h> nested inside.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---

diff --git a/arch/x86/x86_intrins.h b/arch/x86/x86_intrins.h
index a28e3cf74..4f6b7540d 100644
--- a/arch/x86/x86_intrins.h
+++ b/arch/x86/x86_intrins.h
@@ -100,25 +100,15 @@ static inline __m512i _mm512_zextsi128_si512(__m128i a) {
  * corruption. _mm_loadl_epi64 compiles to a single MOVQ xmm,m64 that
  * bypasses the buggy synthesis path. Fixed in VS 2022 17.11 (v143).
  *
- * MSVC only defines _mm_cvtsi64_si128 / _mm_cvtsi128_si64 on 64-bit, so the
- * 32-bit polyfills below are also needed for basic compilation.
+ * GCC and MSVC only define _mm_cvtsi64_si128 / _mm_cvtsi128_si64 on 64-bit,
+ * so the 32-bit polyfills below are also needed for basic compilation.
  *
  * https://developercommunity.visualstudio.com/t/10853479
  */
-#if defined(_MSC_VER) && !defined(__clang__) && defined(ARCH_32BIT)
+#if !defined(__clang__) && defined(ARCH_32BIT)
+#ifdef _MSC_VER
 #include <intrin.h>
-static inline int64_t _mm_cvtsi128_si64(__m128i a) {
-    union { __m128i v; int64_t i; } u;
-    u.v = a;
-    return u.i;
-}
-
-static inline __m128i _mm_cvtsi64_si128(int64_t a) {
-    return _mm_loadl_epi64((const __m128i*)&a);
-}
 #endif
-
-#if defined(__GNUC__) && !defined(__clang__) && defined(ARCH_32BIT)
 static inline int64_t _mm_cvtsi128_si64(__m128i a) {
     union { __m128i v; int64_t i; } u;
     u.v = a;