diff --git a/include/llvm/IntrinsicsX86.td b/include/llvm/IntrinsicsX86.td
index 3ca9cb44419..cea48568108 100644
--- a/include/llvm/IntrinsicsX86.td
+++ b/include/llvm/IntrinsicsX86.td
@@ -786,9 +786,9 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_aesni_aesdeclast : GCCBuiltin<"__builtin_ia32_aesdeclast128">,
               Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty],
                         [IntrNoMem]>;
-  def int_x86_aesni_aeskeygenassist : 
+  def int_x86_aesni_aeskeygenassist :
               GCCBuiltin<"__builtin_ia32_aeskeygenassist128">,
-              Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_i32_ty],
+              Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_i8_ty],
                         [IntrNoMem]>;
 }
 
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 7c69dcd0267..dcd043c73a1 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -117,17 +117,17 @@ def alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
   return cast<LoadSDNode>(N)->getAlignment() >= 16;
 }]>;
 
-def alignedloadfsf32 : PatFrag<(ops node:$ptr), 
+def alignedloadfsf32 : PatFrag<(ops node:$ptr),
                                (f32 (alignedload node:$ptr))>;
-def alignedloadfsf64 : PatFrag<(ops node:$ptr), 
+def alignedloadfsf64 : PatFrag<(ops node:$ptr),
                                (f64 (alignedload node:$ptr))>;
-def alignedloadv4f32 : PatFrag<(ops node:$ptr), 
+def alignedloadv4f32 : PatFrag<(ops node:$ptr),
                                (v4f32 (alignedload node:$ptr))>;
-def alignedloadv2f64 : PatFrag<(ops node:$ptr), 
+def alignedloadv2f64 : PatFrag<(ops node:$ptr),
                                (v2f64 (alignedload node:$ptr))>;
-def alignedloadv4i32 : PatFrag<(ops node:$ptr), 
+def alignedloadv4i32 : PatFrag<(ops node:$ptr),
                                (v4i32 (alignedload node:$ptr))>;
-def alignedloadv2i64 : PatFrag<(ops node:$ptr), 
+def alignedloadv2i64 : PatFrag<(ops node:$ptr),
                                (v2i64 (alignedload node:$ptr))>;
 
 // Like 'load', but uses special alignment checks suitable for use in
@@ -518,25 +518,25 @@ def UCOMISSrr: PSI<0x2E, MRMSrcReg, (outs), (ins FR32:$src1, FR32:$src2),
 def UCOMISSrm: PSI<0x2E, MRMSrcMem, (outs), (ins FR32:$src1, f32mem:$src2),
                    "ucomiss\t{$src2, $src1|$src1, $src2}",
                    [(set EFLAGS, (X86cmp FR32:$src1, (loadf32 addr:$src2)))]>;
-                    
+
 def COMISSrr: PSI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
                   "comiss\t{$src2, $src1|$src1, $src2}", []>;
 def COMISSrm: PSI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
                   "comiss\t{$src2, $src1|$src1, $src2}", []>;
-                  
+
 } // Defs = [EFLAGS]
 
 // Aliases to match intrinsics which expect XMM operand(s).
 let Constraints = "$src1 = $dst" in {
   def Int_CMPSSrr : SSIi8<0xC2, MRMSrcReg,
-                        (outs VR128:$dst), 
+                        (outs VR128:$dst),
                         (ins VR128:$src1, VR128:$src, SSECC:$cc),
                         "cmp${cc}ss\t{$src, $dst|$dst, $src}",
-                        [(set VR128:$dst, (int_x86_sse_cmp_ss 
+                        [(set VR128:$dst, (int_x86_sse_cmp_ss
                                              VR128:$src1,
                                              VR128:$src, imm:$cc))]>;
   def Int_CMPSSrm : SSIi8<0xC2, MRMSrcMem,
-                        (outs VR128:$dst), 
+                        (outs VR128:$dst),
                         (ins VR128:$src1, f32mem:$src, SSECC:$cc),
                         "cmp${cc}ss\t{$src, $dst|$dst, $src}",
                         [(set VR128:$dst, (int_x86_sse_cmp_ss VR128:$src1,
@@ -1312,13 +1312,13 @@ def UCOMISDrm: PDI<0x2E, MRMSrcMem, (outs), (ins FR64:$src1, f64mem:$src2),
 // Aliases to match intrinsics which expect XMM operand(s).
 let Constraints = "$src1 = $dst" in {
   def Int_CMPSDrr : SDIi8<0xC2, MRMSrcReg,
-                        (outs VR128:$dst), 
+                        (outs VR128:$dst),
                         (ins VR128:$src1, VR128:$src, SSECC:$cc),
                         "cmp${cc}sd\t{$src, $dst|$dst, $src}",
                         [(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
                                            VR128:$src, imm:$cc))]>;
   def Int_CMPSDrm : SDIi8<0xC2, MRMSrcMem,
-                        (outs VR128:$dst), 
+                        (outs VR128:$dst),
                         (ins VR128:$src1, f64mem:$src, SSECC:$cc),
                         "cmp${cc}sd\t{$src, $dst|$dst, $src}",
                         [(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
@@ -1656,7 +1656,7 @@ def CVTTPS2DQrm : SSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
 
 def Int_CVTTPS2DQrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                         "cvttps2dq\t{$src, $dst|$dst, $src}",
-                        [(set VR128:$dst, 
+                        [(set VR128:$dst,
                               (int_x86_sse2_cvttps2dq VR128:$src))]>,
                       XS, Requires<[HasSSE2]>;
 def Int_CVTTPS2DQrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
@@ -1981,24 +1981,24 @@ let Constraints = "$src1 = $dst" in {
 
 multiclass PDI_binop_rm_int<bits<8> opc, string OpcodeStr, Intrinsic IntId,
                             bit Commutable = 0> {
-  def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), 
+  def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
                                (ins VR128:$src1, VR128:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (IntId VR128:$src1, VR128:$src2))]> {
     let isCommutable = Commutable;
   }
-  def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), 
+  def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst),
                                (ins VR128:$src1, i128mem:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (IntId VR128:$src1,
-                                        (bitconvert (memopv2i64 
+                                        (bitconvert (memopv2i64
                                                      addr:$src2))))]>;
 }
 
 multiclass PDI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
                              string OpcodeStr,
                              Intrinsic IntId, Intrinsic IntId2> {
-  def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), 
+  def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
                                (ins VR128:$src1, VR128:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (IntId VR128:$src1, VR128:$src2))]>;
@@ -2007,7 +2007,7 @@ multiclass PDI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (IntId VR128:$src1,
                                       (bitconvert (memopv2i64 addr:$src2))))]>;
-  def ri : PDIi8<opc2, ImmForm, (outs VR128:$dst), 
+  def ri : PDIi8<opc2, ImmForm, (outs VR128:$dst),
                                 (ins VR128:$src1, i32i8imm:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (IntId2 VR128:$src1, (i32 imm:$src2)))]>;
@@ -2016,13 +2016,13 @@ multiclass PDI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
 /// PDI_binop_rm - Simple SSE2 binary operator.
 multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
                         ValueType OpVT, bit Commutable = 0> {
-  def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), 
+  def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
                                (ins VR128:$src1, VR128:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (OpVT (OpNode VR128:$src1, VR128:$src2)))]> {
     let isCommutable = Commutable;
   }
-  def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), 
+  def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst),
                                (ins VR128:$src1, i128mem:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (OpVT (OpNode VR128:$src1,
@@ -2908,7 +2908,7 @@ defm PMADDUBSW   : SS3I_binop_rm_int_8 <0x04, "pmaddubsw",
 defm PMULHRSW    : SS3I_binop_rm_int_16<0x0B, "pmulhrsw",
                                         int_x86_ssse3_pmul_hr_sw,
                                         int_x86_ssse3_pmul_hr_sw_128, 1>;
-                                        
+
 defm PSHUFB      : SS3I_binop_rm_int_8 <0x00, "pshufb",
                                         int_x86_ssse3_pshuf_b,
                                         int_x86_ssse3_pshuf_b_128>;
@@ -3469,14 +3469,14 @@ let Constraints = "$src1 = $dst" in {
 let Constraints = "$src1 = $dst" in {
 multiclass SS48I_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
                         ValueType OpVT, bit Commutable = 0> {
-  def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), 
+  def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst),
                                  (ins VR128:$src1, VR128:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (OpVT (OpNode VR128:$src1, VR128:$src2)))]>,
                OpSize {
     let isCommutable = Commutable;
   }
-  def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), 
+  def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst),
                                  (ins VR128:$src1, i128mem:$src2),
                !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
                [(set VR128:$dst, (OpNode VR128:$src1,
@@ -3954,15 +3954,15 @@ let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
 def PCMPESTRM128REG : SS42AI<0, Pseudo, (outs VR128:$dst),
   (ins VR128:$src1, VR128:$src3, i8imm:$src5),
   "#PCMPESTRM128rr PSEUDO!",
-  [(set VR128:$dst, 
-        (int_x86_sse42_pcmpestrm128 
+  [(set VR128:$dst,
+        (int_x86_sse42_pcmpestrm128
          VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5))]>, OpSize;
 
 def PCMPESTRM128MEM : SS42AI<0, Pseudo, (outs VR128:$dst),
   (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
   "#PCMPESTRM128rm PSEUDO!",
-  [(set VR128:$dst, (int_x86_sse42_pcmpestrm128 
-                     VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5))]>, 
+  [(set VR128:$dst, (int_x86_sse42_pcmpestrm128
+                     VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5))]>,
   OpSize;
 }
 
@@ -3977,7 +3977,7 @@ def PCMPESTRM128rm : SS42AI<0x60, MRMSrcMem, (outs),
 
 let Defs = [ECX, EFLAGS] in {
   multiclass SS42AI_pcmpistri<Intrinsic IntId128> {
-    def rr : SS42AI<0x63, MRMSrcReg, (outs), 
+    def rr : SS42AI<0x63, MRMSrcReg, (outs),
       (ins VR128:$src1, VR128:$src2, i8imm:$src3),
       "pcmpistri\t{$src3, $src2, $src1|$src1, $src2, $src3}",
       [(set ECX, (IntId128 VR128:$src1, VR128:$src2, imm:$src3)),
@@ -4008,7 +4008,7 @@ let Uses = [EAX, EDX] in {
     def rm : SS42AI<0x61, MRMSrcMem, (outs),
       (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
        "pcmpestri\t{$src5, $src3, $src1|$src1, $src3, $src5}",
-       [(set ECX, 
+       [(set ECX,
              (IntId128 VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5)),
         (implicit EFLAGS)]>, OpSize;
   }
@@ -4086,16 +4086,15 @@ def AESIMCrm : AES8I<0xDB, MRMSrcMem, (outs VR128:$dst),
   OpSize;
 
 def AESKEYGENASSIST128rr : AESAI<0xDF, MRMSrcReg, (outs VR128:$dst),
-  (ins VR128:$src1, i32i8imm:$src2),
+  (ins VR128:$src1, i8imm:$src2),
   "aeskeygenassist\t{$src2, $src1, $dst|$dst, $src1, $src2}",
   [(set VR128:$dst,
     (int_x86_aesni_aeskeygenassist VR128:$src1, imm:$src2))]>,
   OpSize;
 def AESKEYGENASSIST128rm : AESAI<0xDF, MRMSrcMem, (outs VR128:$dst),
-  (ins i128mem:$src1, i32i8imm:$src2),
+  (ins i128mem:$src1, i8imm:$src2),
   "aeskeygenassist\t{$src2, $src1, $dst|$dst, $src1, $src2}",
   [(set VR128:$dst,
     (int_x86_aesni_aeskeygenassist (bitconvert (memopv2i64 addr:$src1)),
                                     imm:$src2))]>,
   OpSize;
-
diff --git a/test/MC/AsmParser/X86/x86_32-encoding.s b/test/MC/AsmParser/X86/x86_32-encoding.s
index f973cb23b57..994dbb85990 100644
--- a/test/MC/AsmParser/X86/x86_32-encoding.s
+++ b/test/MC/AsmParser/X86/x86_32-encoding.s
@@ -9962,11 +9962,16 @@
 // CHECK:  encoding: [0x66,0x0f,0x3a,0xdf,0x14,0x82,0x7d]
                 aeskeygenassist $125, (%edx,%eax,4), %xmm2
 
+// rdar://8017638
+// CHECK: aeskeygenassist	$128, %xmm1, %xmm2
+// CHECK:  encoding: [0x66,0x0f,0x3a,0xdf,0x14,0x82,0x80]
+		aeskeygenassist $128, %xmm1, %xmm2
+
 // rdar://7840289
 // CHECK: pshufb	CPI1_0(%rip), %xmm1
 // CHECK:  encoding: [0x66,0x0f,0x38,0x00,0x0d,A,A,A,A]
 // CHECK:  fixup A - offset: 5, value: CPI1_0-4
-pshufb	CPI1_0(%rip), %xmm1   
+pshufb	CPI1_0(%rip), %xmm1
 
 // rdar://7910087
 // CHECK: bsfw	%bx, %bx
@@ -10018,7 +10023,7 @@ pshufb	CPI1_0(%rip), %xmm1
 // radr://7914715
 // CHECK: fcoml   3735928559(%ebx,%ecx,8)
 // CHECK:  encoding: [0xdc,0x94,0xcb,0xef,0xbe,0xad,0xde]
-          fcoml   3735928559(%ebx,%ecx,8) 
+          fcoml   3735928559(%ebx,%ecx,8)
 
 // CHECK: fcoms   32493
 // CHECK:  encoding: [0xd8,0x15,0xed,0x7e,0x00,0x00]