diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14905,7 +14905,7 @@
 
     Value *Call = Builder.CreateCall(CGM.getIntrinsic(IID), {Ops[0], Ops[1]});
 
-    for (int i = 0; i < 6; ++i) {
+    for (int i = 0; i < 3; ++i) {
       Value *Extract = Builder.CreateExtractValue(Call, i + 1);
       Value *Ptr = Builder.CreateConstGEP1_32(Int8Ty, Ops[2], i * 16);
       Ptr = Builder.CreateBitCast(
@@ -14921,7 +14921,7 @@
     Value *Call =
         Builder.CreateCall(CGM.getIntrinsic(IID), {Ops[0], Ops[1], Ops[2]});
 
-    for (int i = 0; i < 7; ++i) {
+    for (int i = 0; i < 4; ++i) {
       Value *Extract = Builder.CreateExtractValue(Call, i + 1);
       Value *Ptr = Builder.CreateConstGEP1_32(Int8Ty, Ops[3], i * 16);
       Ptr = Builder.CreateBitCast(
diff --git a/clang/lib/Headers/keylockerintrin.h b/clang/lib/Headers/keylockerintrin.h
--- a/clang/lib/Headers/keylockerintrin.h
+++ b/clang/lib/Headers/keylockerintrin.h
@@ -99,7 +99,7 @@
 }
 
 /// Wrap a 128-bit AES key from __key into a key handle and output in
-/// ((__m128i*)__h) to ((__m128i*)__h) + 5  and a 32-bit value as return.
+/// ((__m128i*)__h) to ((__m128i*)__h) + 2  and a 32-bit value as return.
 /// The explicit source operand __htype specifies handle restrictions.
 ///
 /// \headerfile <x86intrin.h>
@@ -120,9 +120,6 @@
 /// MEM[__h+127:__h] := Handle[127:0]   // AAD
 /// MEM[__h+255:__h+128] := Handle[255:128] // Integrity Tag
 /// MEM[__h+383:__h+256] := Handle[383:256] // CipherText
-/// MEM[__h+511:__h+384] := 0 // Reserved for future usage
-/// MEM[__h+639:__h+512] := 0 // Reserved for future usage
-/// MEM[__h+767:__h+640] := 0 // Reserved for future usage
 /// OF := 0
 /// SF := 0
 /// ZF := 0
@@ -136,7 +133,7 @@
 }
 
 /// Wrap a 256-bit AES key from __key_hi:__key_lo into a key handle, then
-/// output handle in ((__m128i*)__h) to ((__m128i*)__h) + 6 and
+/// output handle in ((__m128i*)__h) to ((__m128i*)__h) + 3 and
 /// a 32-bit value as return.
 /// The explicit source operand __htype specifies handle restrictions.
 ///
@@ -160,9 +157,6 @@
 /// MEM[__h+255:__h+128] := Handle[255:128] // Tag
 /// MEM[__h+383:__h+256] := Handle[383:256] // CipherText[127:0]
 /// MEM[__h+511:__h+384] := Handle[511:384] // CipherText[255:128]
-/// MEM[__h+639:__h+512] := 0 // Reserved for future usage
-/// MEM[__h+767:__h+640] := 0 // Reserved for future usage
-/// MEM[__h+895:__h+768] := 0 Integrity// Reserved for future usage
 /// OF := 0
 /// SF := 0
 /// ZF := 0
diff --git a/clang/test/CodeGen/X86/keylocker.c b/clang/test/CodeGen/X86/keylocker.c
--- a/clang/test/CodeGen/X86/keylocker.c
+++ b/clang/test/CodeGen/X86/keylocker.c
@@ -98,20 +98,8 @@
 // CHECK64-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8* [[TMP5]], i32 32
 // CHECK64-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to <2 x i64>*
 // CHECK64-NEXT:    store <2 x i64> [[TMP12]], <2 x i64>* [[TMP14]], align 1
-// CHECK64-NEXT:    [[TMP15:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 4
-// CHECK64-NEXT:    [[TMP16:%.*]] = getelementptr i8, i8* [[TMP5]], i32 48
-// CHECK64-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to <2 x i64>*
-// CHECK64-NEXT:    store <2 x i64> [[TMP15]], <2 x i64>* [[TMP17]], align 1
-// CHECK64-NEXT:    [[TMP18:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 5
-// CHECK64-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[TMP5]], i32 64
-// CHECK64-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to <2 x i64>*
-// CHECK64-NEXT:    store <2 x i64> [[TMP18]], <2 x i64>* [[TMP20]], align 1
-// CHECK64-NEXT:    [[TMP21:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 6
-// CHECK64-NEXT:    [[TMP22:%.*]] = getelementptr i8, i8* [[TMP5]], i32 80
-// CHECK64-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP22]] to <2 x i64>*
-// CHECK64-NEXT:    store <2 x i64> [[TMP21]], <2 x i64>* [[TMP23]], align 1
-// CHECK64-NEXT:    [[TMP24:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 0
-// CHECK64-NEXT:    ret i32 [[TMP24]]
+// CHECK64-NEXT:    [[TMP15:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 0
+// CHECK64-NEXT:    ret i32 [[TMP15]]
 //
 // CHECK32-LABEL: @test_encodekey128_u32(
 // CHECK32-NEXT:  entry:
@@ -145,20 +133,8 @@
 // CHECK32-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8* [[TMP5]], i32 32
 // CHECK32-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to <2 x i64>*
 // CHECK32-NEXT:    store <2 x i64> [[TMP12]], <2 x i64>* [[TMP14]], align 1
-// CHECK32-NEXT:    [[TMP15:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 4
-// CHECK32-NEXT:    [[TMP16:%.*]] = getelementptr i8, i8* [[TMP5]], i32 48
-// CHECK32-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to <2 x i64>*
-// CHECK32-NEXT:    store <2 x i64> [[TMP15]], <2 x i64>* [[TMP17]], align 1
-// CHECK32-NEXT:    [[TMP18:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 5
-// CHECK32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[TMP5]], i32 64
-// CHECK32-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to <2 x i64>*
-// CHECK32-NEXT:    store <2 x i64> [[TMP18]], <2 x i64>* [[TMP20]], align 1
-// CHECK32-NEXT:    [[TMP21:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 6
-// CHECK32-NEXT:    [[TMP22:%.*]] = getelementptr i8, i8* [[TMP5]], i32 80
-// CHECK32-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP22]] to <2 x i64>*
-// CHECK32-NEXT:    store <2 x i64> [[TMP21]], <2 x i64>* [[TMP23]], align 1
-// CHECK32-NEXT:    [[TMP24:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 0
-// CHECK32-NEXT:    ret i32 [[TMP24]]
+// CHECK32-NEXT:    [[TMP15:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP6]], 0
+// CHECK32-NEXT:    ret i32 [[TMP15]]
 //
 unsigned int test_encodekey128_u32(unsigned int htype, __m128i key, void *h) {
   return _mm_encodekey128_u32(htype, key, h);
@@ -206,20 +182,8 @@
 // CHECK64-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP7]], i32 48
 // CHECK64-NEXT:    [[TMP19:%.*]] = bitcast i8* [[TMP18]] to <2 x i64>*
 // CHECK64-NEXT:    store <2 x i64> [[TMP17]], <2 x i64>* [[TMP19]], align 1
-// CHECK64-NEXT:    [[TMP20:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 5
-// CHECK64-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP7]], i32 64
-// CHECK64-NEXT:    [[TMP22:%.*]] = bitcast i8* [[TMP21]] to <2 x i64>*
-// CHECK64-NEXT:    store <2 x i64> [[TMP20]], <2 x i64>* [[TMP22]], align 1
-// CHECK64-NEXT:    [[TMP23:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 6
-// CHECK64-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[TMP7]], i32 80
-// CHECK64-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to <2 x i64>*
-// CHECK64-NEXT:    store <2 x i64> [[TMP23]], <2 x i64>* [[TMP25]], align 1
-// CHECK64-NEXT:    [[TMP26:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 7
-// CHECK64-NEXT:    [[TMP27:%.*]] = getelementptr i8, i8* [[TMP7]], i32 96
-// CHECK64-NEXT:    [[TMP28:%.*]] = bitcast i8* [[TMP27]] to <2 x i64>*
-// CHECK64-NEXT:    store <2 x i64> [[TMP26]], <2 x i64>* [[TMP28]], align 1
-// CHECK64-NEXT:    [[TMP29:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 0
-// CHECK64-NEXT:    ret i32 [[TMP29]]
+// CHECK64-NEXT:    [[TMP20:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 0
+// CHECK64-NEXT:    ret i32 [[TMP20]]
 //
 // CHECK32-LABEL: @test_encodekey256_u32(
 // CHECK32-NEXT:  entry:
@@ -263,20 +227,8 @@
 // CHECK32-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[TMP7]], i32 48
 // CHECK32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[TMP18]] to <2 x i64>*
 // CHECK32-NEXT:    store <2 x i64> [[TMP17]], <2 x i64>* [[TMP19]], align 1
-// CHECK32-NEXT:    [[TMP20:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 5
-// CHECK32-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP7]], i32 64
-// CHECK32-NEXT:    [[TMP22:%.*]] = bitcast i8* [[TMP21]] to <2 x i64>*
-// CHECK32-NEXT:    store <2 x i64> [[TMP20]], <2 x i64>* [[TMP22]], align 1
-// CHECK32-NEXT:    [[TMP23:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 6
-// CHECK32-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[TMP7]], i32 80
-// CHECK32-NEXT:    [[TMP25:%.*]] = bitcast i8* [[TMP24]] to <2 x i64>*
-// CHECK32-NEXT:    store <2 x i64> [[TMP23]], <2 x i64>* [[TMP25]], align 1
-// CHECK32-NEXT:    [[TMP26:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 7
-// CHECK32-NEXT:    [[TMP27:%.*]] = getelementptr i8, i8* [[TMP7]], i32 96
-// CHECK32-NEXT:    [[TMP28:%.*]] = bitcast i8* [[TMP27]] to <2 x i64>*
-// CHECK32-NEXT:    store <2 x i64> [[TMP26]], <2 x i64>* [[TMP28]], align 1
-// CHECK32-NEXT:    [[TMP29:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 0
-// CHECK32-NEXT:    ret i32 [[TMP29]]
+// CHECK32-NEXT:    [[TMP20:%.*]] = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[TMP8]], 0
+// CHECK32-NEXT:    ret i32 [[TMP20]]
 //
 unsigned int test_encodekey256_u32(unsigned int htype, __m128i key_lo, __m128i key_hi, void *h) {
   return _mm_encodekey256_u32(htype, key_lo, key_hi, h);
diff --git a/llvm/test/CodeGen/X86/keylocker-intrinsics.ll b/llvm/test/CodeGen/X86/keylocker-intrinsics.ll
--- a/llvm/test/CodeGen/X86/keylocker-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/keylocker-intrinsics.ll
@@ -36,40 +36,24 @@
 define i32 @test_encodekey128_u32(i32 %htype, <2 x i64> %key, <2 x i64>* nocapture %h0, <2 x i64>* nocapture %h1, <2 x i64>* nocapture %h2, <2 x i64>* nocapture %h3, <2 x i64>* nocapture %h4, <2 x i64>* nocapture %h5) nounwind {
 ; X64-LABEL: test_encodekey128_u32:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; X64-NEXT:    encodekey128 %edi, %eax
 ; X64-NEXT:    movaps %xmm0, (%rsi)
 ; X64-NEXT:    movaps %xmm1, (%rdx)
 ; X64-NEXT:    movaps %xmm2, (%rcx)
-; X64-NEXT:    movaps %xmm4, (%r8)
-; X64-NEXT:    movaps %xmm5, (%r9)
-; X64-NEXT:    movaps %xmm6, (%r10)
 ; X64-NEXT:    retq
 ;
 ; X32-LABEL: test_encodekey128_u32:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    pushl %ebp
-; X32-NEXT:    pushl %ebx
-; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %esi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    encodekey128 %eax, %eax
-; X32-NEXT:    vmovaps %xmm0, (%ebp)
-; X32-NEXT:    vmovaps %xmm1, (%ebx)
-; X32-NEXT:    vmovaps %xmm2, (%edi)
-; X32-NEXT:    vmovaps %xmm4, (%esi)
-; X32-NEXT:    vmovaps %xmm5, (%edx)
-; X32-NEXT:    vmovaps %xmm6, (%ecx)
+; X32-NEXT:    vmovaps %xmm0, (%esi)
+; X32-NEXT:    vmovaps %xmm1, (%edx)
+; X32-NEXT:    vmovaps %xmm2, (%ecx)
 ; X32-NEXT:    popl %esi
-; X32-NEXT:    popl %edi
-; X32-NEXT:    popl %ebx
-; X32-NEXT:    popl %ebp
 ; X32-NEXT:    retl
 entry:
   %0 = tail call { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.x86.encodekey128(i32 %htype, <2 x i64> %key)
@@ -79,53 +63,36 @@
   store <2 x i64> %2, <2 x i64>* %h1, align 16
   %3 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 3
   store <2 x i64> %3, <2 x i64>* %h2, align 16
-  %4 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 4
-  store <2 x i64> %4, <2 x i64>* %h3, align 16
-  %5 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 5
-  store <2 x i64> %5, <2 x i64>* %h4, align 16
-  %6 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 6
-  store <2 x i64> %6, <2 x i64>* %h5, align 16
-  %7 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 0
-  ret i32 %7
+  %4 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 0
+  ret i32 %4
 }
 
 define i32 @test_encodekey256_u32(i32 %htype, <2 x i64> %key_lo, <2 x i64> %key_hi, <2 x i64>* nocapture %h0, <2 x i64>* nocapture %h1, <2 x i64>* nocapture %h2, <2 x i64>* nocapture %h3, <2 x i64>* nocapture %h4, <2 x i64>* nocapture %h5, <2 x      i64>* nocapture readnone %h6) nounwind {
 ; X64-LABEL: test_encodekey256_u32:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; X64-NEXT:    encodekey256 %edi, %eax
 ; X64-NEXT:    movaps %xmm0, (%rsi)
 ; X64-NEXT:    movaps %xmm1, (%rdx)
 ; X64-NEXT:    movaps %xmm2, (%rcx)
 ; X64-NEXT:    movaps %xmm3, (%r8)
-; X64-NEXT:    movaps %xmm4, (%r9)
-; X64-NEXT:    movaps %xmm5, (%r10)
 ; X64-NEXT:    retq
 ;
 ; X32-LABEL: test_encodekey256_u32:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    pushl %ebp
-; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    pushl %edi
 ; X32-NEXT:    pushl %esi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    encodekey256 %eax, %eax
-; X32-NEXT:    vmovaps %xmm0, (%ebp)
-; X32-NEXT:    vmovaps %xmm1, (%ebx)
-; X32-NEXT:    vmovaps %xmm2, (%edi)
-; X32-NEXT:    vmovaps %xmm3, (%esi)
-; X32-NEXT:    vmovaps %xmm4, (%edx)
-; X32-NEXT:    vmovaps %xmm5, (%ecx)
+; X32-NEXT:    vmovaps %xmm0, (%edi)
+; X32-NEXT:    vmovaps %xmm1, (%esi)
+; X32-NEXT:    vmovaps %xmm2, (%edx)
+; X32-NEXT:    vmovaps %xmm3, (%ecx)
 ; X32-NEXT:    popl %esi
 ; X32-NEXT:    popl %edi
-; X32-NEXT:    popl %ebx
-; X32-NEXT:    popl %ebp
 ; X32-NEXT:    retl
 entry:
   %0 = tail call { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.x86.encodekey256(i32 %htype, <2 x i64> %key_lo, <2 x i64> %key_hi)
@@ -137,12 +104,8 @@
   store <2 x i64> %3, <2 x i64>* %h2, align 16
   %4 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 4
   store <2 x i64> %4, <2 x i64>* %h3, align 16
-  %5 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 5
-  store <2 x i64> %5, <2 x i64>* %h4, align 16
-  %6 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 6
-  store <2 x i64> %6, <2 x i64>* %h5, align 16
-  %7 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 0
-  ret i32 %7
+  %5 = extractvalue { i32, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %0, 0
+  ret i32 %5
 }
 
 define i8 @test_mm_aesenc128kl_u8(<2 x i64> %data, i8* %h, <2 x i64>* %out) {