[X86] Move back _mulx_u32 to 32-bit only

[pengfei]

Summary

We cannot lower it to mulx at the present due to backend reason.
Since GCC doesn't support it on 64-bit, it's OK to leave it 32-bit only.

[probinson]

I'm okay with putting this back if the codegen (D153620) can't be resolved. Honestly I didn't think it was a problem.
But I think Craig or Simon should sign off.

clang/lib/Headers/bmi2intrin.h

 /*===---- bmi2intrin.h - BMI2 intrinsics -----------------------------------===
  *
  * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
  * See https://llvm.org/LICENSE.txt for license information.
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  *
  *===-----------------------------------------------------------------------===
  */

 }
 
 static __inline__ unsigned int __DEFAULT_FN_ATTRS
 _pext_u32(unsigned int __X, unsigned int __Y)
 {
   return __builtin_ia32_pext_si(__X, __Y);
 }
 
-static __inline__ unsigned int __DEFAULT_FN_ATTRS
-_mulx_u32(unsigned int __X, unsigned int __Y, unsigned int *__P)
-{
-  unsigned long long __res = (unsigned long long) __X * __Y;
-  *__P = (unsigned int)(__res >> 32);
-  return (unsigned int)__res;
-}
-
 #ifdef  __x86_64__
 
 static __inline__ unsigned long long __DEFAULT_FN_ATTRS
 _bzhi_u64(unsigned long long __X, unsigned long long __Y)
 {
   return __builtin_ia32_bzhi_di(__X, __Y);
 }
 

 _mulx_u64 (unsigned long long __X, unsigned long long __Y,
      unsigned long long *__P)
 {
   unsigned __int128 __res = (unsigned __int128) __X * __Y;
   *__P = (unsigned long long) (__res >> 64);
   return (unsigned long long) __res;
 }
 
-#endif /* __x86_64__  */
+#else /* !__x86_64__ */
+
+static __inline__ unsigned int __DEFAULT_FN_ATTRS
+_mulx_u32 (unsigned int __X, unsigned int __Y, unsigned int *__P)
+{
+  unsigned long long __res = (unsigned long long) __X * __Y;
+  *__P = (unsigned int) (__res >> 32);
+  return (unsigned int) __res;
+}
+
+#endif /* !__x86_64__ */
 
 #undef __DEFAULT_FN_ATTRS
 
 #endif /* __BMI2INTRIN_H */