diff --git a/lib/Headers/bmi2intrin.h b/lib/Headers/bmi2intrin.h
index c60b0c4..a05cfad 100644
--- a/lib/Headers/bmi2intrin.h
+++ b/lib/Headers/bmi2intrin.h
@@ -70,6 +70,25 @@
   return __builtin_ia32_pext_di(__X, __Y);
 }
 
+static __inline__ unsigned long long __attribute__((__always_inline__, __nodebug__))
+_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;
+}
+
+#else /* !__x86_64__ */
+
+static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
+_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__  */
 
 #endif /* __BMI2INTRIN_H */
diff --git a/test/CodeGen/bmi2-builtins.c b/test/CodeGen/bmi2-builtins.c
index 18b2319..201cac6 100644
--- a/test/CodeGen/bmi2-builtins.c
+++ b/test/CodeGen/bmi2-builtins.c
@@ -1,4 +1,5 @@
 // RUN: %clang_cc1 %s -O3 -triple=x86_64-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 %s -O3 -triple=i386-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s --check-prefix=B32
 
 // Don't include mm_malloc.h, it's system specific.
 #define __MM_MALLOC_H
@@ -20,6 +21,15 @@
   return _pext_u32(__X, __Y);
 }
 
+unsigned int test_mulx_u32(unsigned int __X, unsigned int __Y,
+                                 unsigned int *__P) {
+  // CHECK: @test_mulx_u32
+  // CHECK-NOT: mul i64
+  // B32: @test_mulx_u32
+  // B32: mul i64
+  return _mulx_u32(__X, __Y, __P);
+}
+
 unsigned long long test_bzhi_u64(unsigned long long __X, unsigned long long __Y) {
   // CHECK: @llvm.x86.bmi.bzhi.64
   return _bzhi_u64(__X, __Y);
@@ -34,3 +44,10 @@
   // CHECK: @llvm.x86.bmi.pext.64
   return _pext_u64(__X, __Y);
 }
+
+unsigned long long test_mulx_u64(unsigned long long __X, unsigned long long __Y,
+                                 unsigned long long *__P) {
+  // CHECK: @test_mulx_u64
+  // CHECK: mul i128
+  return _mulx_u64(__X, __Y, __P);
+}