Introduce 'UseSSEx' to force SSE legacy encoding

- Add 'UseSSEx' to force SSE legacy insn not being selected when AVX is
  enabled.

  As the penalty of inter-mixing SSE and AVX instructions, we need
  prevent SSE legacy insn from being generated except explicitly
  specified through some intrinsics. For patterns supported by both
  SSE and AVX, so far, we force AVX insn will be tried first relying on
  AddedComplexity or position in td file. It's error-prone and
  introduces bugs accidentally.

  'UseSSEx' is disabled when AVX is turned on. For SSE insns inherited
  by AVX, we need this predicate to force VEX encoding or SSE legacy
  encoding only.

  For insns not inherited by AVX, we still use the previous predicates,
  i.e. 'HasSSEx'. So far, these insns fall into the following
  categories:
  * SSE insns with MMX operands
  * SSE insns with GPR/MEM operands only (xFENCE, PREFETCH, CLFLUSH,
    CRC, and etc.)
  * SSE4A insns.
  * MMX insns.
  * x87 insns added by SSE.

2 test cases are modified:

 - test/CodeGen/X86/fast-isel-x86-64.ll
   AVX code generation is different from SSE one. 'vcvtsi2sdq' cannot be
   selected by fast-isel due to complicated pattern and fast-isel
   fallback to materialize it from constant pool.

 - test/CodeGen/X86/widen_load-1.ll
   AVX code generation is different from SSE one after fixing SSE/AVX
   inter-mixing. Exec-domain fixing prefers 'vmovapd' instead of
   'vmovaps'.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@162919 91177308-0d34-0410-b5e6-96231b3b80d8

7 files changed, 174 insertions, 127 deletions

diff --git a/lib/Target/X86/X86InstrFormats.td b/lib/Target/X86/X86InstrFormats.td
index 81b4f81..55ad2ec 100644
--- a/lib/Target/X86/X86InstrFormats.td
+++ b/lib/Target/X86/X86InstrFormats.td
@@ -287,12 +287,14 @@ class Iseg32 <bits<8> o, Format f, dag outs, dag ins, string asm,
   let CodeSize = 3;
 }
 
+def __xs : XS;
+
 // SI - SSE 1 & 2 scalar instructions
 class SI<bits<8> o, Format F, dag outs, dag ins, string asm,
          list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin> {
   let Predicates = !if(hasVEXPrefix /* VEX */, [HasAVX],
-            !if(!eq(Prefix, 12 /* XS */), [HasSSE1], [HasSSE2]));
+            !if(!eq(Prefix, __xs.Prefix), [UseSSE1], [UseSSE2]));
 
   // AVX instructions have a 'v' prefix in the mnemonic
   let AsmString = !if(hasVEXPrefix, !strconcat("v", asm), asm);
@@ -303,7 +305,7 @@ class SIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : Ii8<o, F, outs, ins, asm, pattern, itin> {
   let Predicates = !if(hasVEXPrefix /* VEX */, [HasAVX],
-            !if(!eq(Prefix, 12 /* XS */), [HasSSE1], [HasSSE2]));
+            !if(!eq(Prefix, __xs.Prefix), [UseSSE1], [UseSSE2]));
 
   // AVX instructions have a 'v' prefix in the mnemonic
   let AsmString = !if(hasVEXPrefix, !strconcat("v", asm), asm);
@@ -314,18 +316,25 @@ class PI<bits<8> o, Format F, dag outs, dag ins, string asm, list<dag> pattern,
          InstrItinClass itin, Domain d>
       : I<o, F, outs, ins, asm, pattern, itin, d> {
   let Predicates = !if(hasVEXPrefix /* VEX */, [HasAVX],
-        !if(hasOpSizePrefix /* OpSize */, [HasSSE2], [HasSSE1]));
+        !if(hasOpSizePrefix /* OpSize */, [UseSSE2], [UseSSE1]));
 
   // AVX instructions have a 'v' prefix in the mnemonic
   let AsmString = !if(hasVEXPrefix, !strconcat("v", asm), asm);
 }
 
+// MMXPI - SSE 1 & 2 packed instructions with MMX operands
+class MMXPI<bits<8> o, Format F, dag outs, dag ins, string asm, list<dag> pattern,
+            InstrItinClass itin, Domain d>
+      : I<o, F, outs, ins, asm, pattern, itin, d> {
+  let Predicates = !if(hasOpSizePrefix /* OpSize */, [HasSSE2], [HasSSE1]);
+}
+
 // PIi8 - SSE 1 & 2 packed instructions with immediate
 class PIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
            list<dag> pattern, InstrItinClass itin, Domain d>
       : Ii8<o, F, outs, ins, asm, pattern, itin, d> {
   let Predicates = !if(hasVEX_4VPrefix /* VEX */, [HasAVX],
-        !if(hasOpSizePrefix /* OpSize */, [HasSSE2], [HasSSE1]));
+        !if(hasOpSizePrefix /* OpSize */, [UseSSE2], [UseSSE1]));
 
   // AVX instructions have a 'v' prefix in the mnemonic
   let AsmString = !if(hasVEX_4VPrefix, !strconcat("v", asm), asm);
@@ -341,18 +350,18 @@ class PIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
 
 class SSI<bits<8> o, Format F, dag outs, dag ins, string asm,
           list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
-      : I<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[HasSSE1]>;
+      : I<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[UseSSE1]>;
 class SSIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
-      : Ii8<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[HasSSE1]>;
+      : Ii8<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[UseSSE1]>;
 class PSI<bits<8> o, Format F, dag outs, dag ins, string asm,
           list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedSingle>, TB,
-        Requires<[HasSSE1]>;
+        Requires<[UseSSE1]>;
 class PSIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : Ii8<o, F, outs, ins, asm, pattern, itin, SSEPackedSingle>, TB,
-        Requires<[HasSSE1]>;
+        Requires<[UseSSE1]>;
 class VSSI<bits<8> o, Format F, dag outs, dag ins, string asm,
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, !strconcat("v", asm), pattern, itin>, XS,
@@ -372,27 +381,31 @@ class VPSI<bits<8> o, Format F, dag outs, dag ins, string asm,
 //   PDIi8  - SSE2 instructions with ImmT == Imm8 and TB and OpSize prefixes.
 //   VSDI   - SSE2 instructions with XD prefix in AVX form.
 //   VPDI   - SSE2 instructions with TB and OpSize prefixes in AVX form.
+//   MMXSDIi8  - SSE2 instructions with ImmT == Imm8 and XD prefix as well as
+//               MMX operands.
+//   MMXSSDIi8 - SSE2 instructions with ImmT == Imm8 and XS prefix as well as
+//               MMX operands.
 
 class SDI<bits<8> o, Format F, dag outs, dag ins, string asm,
           list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
-      : I<o, F, outs, ins, asm, pattern, itin>, XD, Requires<[HasSSE2]>;
+      : I<o, F, outs, ins, asm, pattern, itin>, XD, Requires<[UseSSE2]>;
 class SDIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
-      : Ii8<o, F, outs, ins, asm, pattern, itin>, XD, Requires<[HasSSE2]>;
+      : Ii8<o, F, outs, ins, asm, pattern, itin>, XD, Requires<[UseSSE2]>;
 class S2SI<bits<8> o, Format F, dag outs, dag ins, string asm,
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
-      : I<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[HasSSE2]>;
+      : I<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[UseSSE2]>;
 class S2SIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
              list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
-      : Ii8<o, F, outs, ins, asm, pattern>, XS, Requires<[HasSSE2]>;
+      : Ii8<o, F, outs, ins, asm, pattern>, XS, Requires<[UseSSE2]>;
 class PDI<bits<8> o, Format F, dag outs, dag ins, string asm,
           list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedDouble>, TB, OpSize,
-        Requires<[HasSSE2]>;
+        Requires<[UseSSE2]>;
 class PDIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : Ii8<o, F, outs, ins, asm, pattern, itin, SSEPackedDouble>, TB, OpSize,
-        Requires<[HasSSE2]>;
+        Requires<[UseSSE2]>;
 class VSDI<bits<8> o, Format F, dag outs, dag ins, string asm,
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, !strconcat("v", asm), pattern, itin>, XD,
@@ -405,6 +418,12 @@ class VPDI<bits<8> o, Format F, dag outs, dag ins, string asm,
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, !strconcat("v", asm), pattern, itin, SSEPackedDouble>, TB,
         OpSize, Requires<[HasAVX]>;
+class MMXSDIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
+               list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
+      : Ii8<o, F, outs, ins, asm, pattern, itin>, XD, Requires<[HasSSE2]>;
+class MMXS2SIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
+                list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
+      : Ii8<o, F, outs, ins, asm, pattern>, XS, Requires<[HasSSE2]>;
 
 // SSE3 Instruction Templates:
 // 
@@ -415,21 +434,23 @@ class VPDI<bits<8> o, Format F, dag outs, dag ins, string asm,
 class S3SI<bits<8> o, Format F, dag outs, dag ins, string asm, 
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedSingle>, XS,
-        Requires<[HasSSE3]>;
+        Requires<[UseSSE3]>;
 class S3DI<bits<8> o, Format F, dag outs, dag ins, string asm, 
            list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedDouble>, XD,
-        Requires<[HasSSE3]>;
+        Requires<[UseSSE3]>;
 class S3I<bits<8> o, Format F, dag outs, dag ins, string asm,
           list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedDouble>, TB, OpSize,
-        Requires<[HasSSE3]>;
+        Requires<[UseSSE3]>;
 
 
 // SSSE3 Instruction Templates:
 // 
 //   SS38I - SSSE3 instructions with T8 prefix.
 //   SS3AI - SSSE3 instructions with TA prefix.
+//   MMXSS38I - SSSE3 instructions with T8 prefix and MMX operands.
+//   MMXSS3AI - SSSE3 instructions with TA prefix and MMX operands.
 //
 // Note: SSSE3 instructions have 64-bit and 128-bit versions. The 64-bit version
 // uses the MMX registers. The 64-bit versions are grouped with the MMX
@@ -438,10 +459,18 @@ class S3I<bits<8> o, Format F, dag outs, dag ins, string asm,
 class SS38I<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, T8,
-        Requires<[HasSSSE3]>;
+        Requires<[UseSSSE3]>;
 class SS3AI<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : Ii8<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, TA,
+        Requires<[UseSSSE3]>;
+class MMXSS38I<bits<8> o, Format F, dag outs, dag ins, string asm,
+               list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
+      : I<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, T8,
+        Requires<[HasSSSE3]>;
+class MMXSS3AI<bits<8> o, Format F, dag outs, dag ins, string asm,
+               list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
+      : Ii8<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, TA,
         Requires<[HasSSSE3]>;
 
 // SSE4.1 Instruction Templates:
@@ -452,11 +481,11 @@ class SS3AI<bits<8> o, Format F, dag outs, dag ins, string asm,
 class SS48I<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, T8,
-        Requires<[HasSSE41]>;
+        Requires<[UseSSE41]>;
 class SS4AIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
             list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : Ii8<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, TA,
-        Requires<[HasSSE41]>;
+        Requires<[UseSSE41]>;
 
 // SSE4.2 Instruction Templates:
 // 
@@ -464,9 +493,10 @@ class SS4AIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
 class SS428I<bits<8> o, Format F, dag outs, dag ins, string asm,
              list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, T8,
-        Requires<[HasSSE42]>;
+        Requires<[UseSSE42]>;
 
 //   SS42FI - SSE 4.2 instructions with T8XD prefix.
+// NOTE: 'HasSSE42' is used as SS42FI is only used for CRC32 insns.
 class SS42FI<bits<8> o, Format F, dag outs, dag ins, string asm,
              list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : I<o, F, outs, ins, asm, pattern, itin>, T8XD, Requires<[HasSSE42]>;
@@ -475,7 +505,7 @@ class SS42FI<bits<8> o, Format F, dag outs, dag ins, string asm,
 class SS42AI<bits<8> o, Format F, dag outs, dag ins, string asm,
              list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
       : Ii8<o, F, outs, ins, asm, pattern, itin, SSEPackedInt>, TA,
-        Requires<[HasSSE42]>;
+        Requires<[UseSSE42]>;
 
 // AVX Instruction Templates:
 //   Instructions introduced in AVX (no SSE equivalent forms)
diff --git a/lib/Target/X86/X86InstrInfo.td b/lib/Target/X86/X86InstrInfo.td
index c8a588b..304676d 100644
--- a/lib/Target/X86/X86InstrInfo.td
+++ b/lib/Target/X86/X86InstrInfo.td
@@ -552,11 +552,17 @@ def HasMMX       : Predicate<"Subtarget->hasMMX()">;
 def Has3DNow     : Predicate<"Subtarget->has3DNow()">;
 def Has3DNowA    : Predicate<"Subtarget->has3DNowA()">;
 def HasSSE1      : Predicate<"Subtarget->hasSSE1()">;
+def UseSSE1      : Predicate<"Subtarget->hasSSE1() && Subtarget->hasNoAVX()">;
 def HasSSE2      : Predicate<"Subtarget->hasSSE2()">;
+def UseSSE2      : Predicate<"Subtarget->hasSSE2() && Subtarget->hasNoAVX()">;
 def HasSSE3      : Predicate<"Subtarget->hasSSE3()">;
+def UseSSE3      : Predicate<"Subtarget->hasSSE3() && Subtarget->hasNoAVX()">;
 def HasSSSE3     : Predicate<"Subtarget->hasSSSE3()">;
+def UseSSSE3     : Predicate<"Subtarget->hasSSSE3() && Subtarget->hasNoAVX()">;
 def HasSSE41     : Predicate<"Subtarget->hasSSE41()">;
+def UseSSE41     : Predicate<"Subtarget->hasSSE41() && Subtarget->hasNoAVX()">;
 def HasSSE42     : Predicate<"Subtarget->hasSSE42()">;
+def UseSSE42     : Predicate<"Subtarget->hasSSE42() && Subtarget->hasNoAVX()">;
 def HasSSE4A     : Predicate<"Subtarget->hasSSE4A()">;
 def HasAVX       : Predicate<"Subtarget->hasAVX()">;
 def HasAVX2      : Predicate<"Subtarget->hasAVX2()">;
diff --git a/lib/Target/X86/X86InstrMMX.td b/lib/Target/X86/X86InstrMMX.td
index c8f40bb..bd54858 100644
--- a/lib/Target/X86/X86InstrMMX.td
+++ b/lib/Target/X86/X86InstrMMX.td
@@ -118,11 +118,11 @@ let Constraints = "$src1 = $dst" in {
 /// Unary MMX instructions requiring SSSE3.
 multiclass SS3I_unop_rm_int_mm<bits<8> opc, string OpcodeStr,
                                Intrinsic IntId64, OpndItins itins> {
-  def rr64 : SS38I<opc, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src),
+  def rr64 : MMXSS38I<opc, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src),
                    !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
                    [(set VR64:$dst, (IntId64 VR64:$src))], itins.rr>;
 
-  def rm64 : SS38I<opc, MRMSrcMem, (outs VR64:$dst), (ins i64mem:$src),
+  def rm64 : MMXSS38I<opc, MRMSrcMem, (outs VR64:$dst), (ins i64mem:$src),
                    !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
                    [(set VR64:$dst,
                      (IntId64 (bitconvert (memopmmx addr:$src))))],
@@ -134,11 +134,11 @@ let ImmT = NoImm, Constraints = "$src1 = $dst" in {
 multiclass SS3I_binop_rm_int_mm<bits<8> opc, string OpcodeStr,
                              Intrinsic IntId64, OpndItins itins> {
   let isCommutable = 0 in
-  def rr64 : SS38I<opc, MRMSrcReg, (outs VR64:$dst),
+  def rr64 : MMXSS38I<opc, MRMSrcReg, (outs VR64:$dst),
        (ins VR64:$src1, VR64:$src2),
         !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
        [(set VR64:$dst, (IntId64 VR64:$src1, VR64:$src2))], itins.rr>;
-  def rm64 : SS38I<opc, MRMSrcMem, (outs VR64:$dst),
+  def rm64 : MMXSS38I<opc, MRMSrcMem, (outs VR64:$dst),
        (ins VR64:$src1, i64mem:$src2),
         !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
        [(set VR64:$dst,
@@ -149,11 +149,11 @@ multiclass SS3I_binop_rm_int_mm<bits<8> opc, string OpcodeStr,
 
 /// PALIGN MMX instructions (require SSSE3).
 multiclass ssse3_palign_mm<string asm, Intrinsic IntId> {
-  def R64irr  : SS3AI<0x0F, MRMSrcReg, (outs VR64:$dst),
+  def R64irr  : MMXSS3AI<0x0F, MRMSrcReg, (outs VR64:$dst),
       (ins VR64:$src1, VR64:$src2, i8imm:$src3),
       !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"), 
       [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2, (i8 imm:$src3)))]>;
-  def R64irm  : SS3AI<0x0F, MRMSrcMem, (outs VR64:$dst),
+  def R64irm  : MMXSS3AI<0x0F, MRMSrcMem, (outs VR64:$dst),
       (ins VR64:$src1, i64mem:$src2, i8imm:$src3),
       !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
       [(set VR64:$dst, (IntId VR64:$src1,
@@ -163,12 +163,10 @@ multiclass ssse3_palign_mm<string asm, Intrinsic IntId> {
 multiclass sse12_cvt_pint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
                          Intrinsic Int, X86MemOperand x86memop, PatFrag ld_frag,
                          string asm, OpndItins itins, Domain d> {
-  def irr : PI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
-                        [(set DstRC:$dst, (Int SrcRC:$src))], 
-                        itins.rr, d>;
-  def irm : PI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
-                        [(set DstRC:$dst, (Int (ld_frag addr:$src)))], 
-                        itins.rm, d>;
+  def irr : MMXPI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
+                  [(set DstRC:$dst, (Int SrcRC:$src))], itins.rr, d>;
+  def irm : MMXPI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
+                  [(set DstRC:$dst, (Int (ld_frag addr:$src)))], itins.rm, d>;
 }
 
 multiclass sse12_cvt_pint_3addr<bits<8> opc, RegisterClass SrcRC,
@@ -243,29 +241,30 @@ def MMX_MOVQ64mr : MMXI<0x7F, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src),
                         [(store (x86mmx VR64:$src), addr:$dst)],
                         IIC_MMX_MOVQ_RM>;
 
-def MMX_MOVDQ2Qrr : SDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
-                          (ins VR128:$src), "movdq2q\t{$src, $dst|$dst, $src}",
-                          [(set VR64:$dst,
-                            (x86mmx (bitconvert
-                            (i64 (vector_extract (v2i64 VR128:$src),
-                                  (iPTR 0))))))],
-                          IIC_MMX_MOVQ_RR>;
-
-def MMX_MOVQ2DQrr : S2SIi8<0xD6, MRMSrcReg, (outs VR128:$dst),
-                            (ins VR64:$src), "movq2dq\t{$src, $dst|$dst, $src}",
-          [(set VR128:$dst,
-            (v2i64 (scalar_to_vector
-                              (i64 (bitconvert (x86mmx VR64:$src))))))],
-                           IIC_MMX_MOVQ_RR>;
+def MMX_MOVDQ2Qrr : MMXSDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
+                             (ins VR128:$src), "movdq2q\t{$src, $dst|$dst, $src}",
+                             [(set VR64:$dst,
+                               (x86mmx (bitconvert
+                               (i64 (vector_extract (v2i64 VR128:$src),
+                                     (iPTR 0))))))],
+                             IIC_MMX_MOVQ_RR>;
+
+def MMX_MOVQ2DQrr : MMXS2SIi8<0xD6, MRMSrcReg, (outs VR128:$dst),
+                              (ins VR64:$src), "movq2dq\t{$src, $dst|$dst, $src}",
+                              [(set VR128:$dst,
+                                (v2i64
+                                  (scalar_to_vector
+                                    (i64 (bitconvert (x86mmx VR64:$src))))))],
+                              IIC_MMX_MOVQ_RR>;
 
 let neverHasSideEffects = 1 in
-def MMX_MOVQ2FR64rr: S2SIi8<0xD6, MRMSrcReg, (outs FR64:$dst),
-                       (ins VR64:$src), "movq2dq\t{$src, $dst|$dst, $src}", [],
-                       IIC_MMX_MOVQ_RR>;
+def MMX_MOVQ2FR64rr: MMXS2SIi8<0xD6, MRMSrcReg, (outs FR64:$dst),
+                               (ins VR64:$src), "movq2dq\t{$src, $dst|$dst, $src}",
+                               [], IIC_MMX_MOVQ_RR>;
 
-def MMX_MOVFR642Qrr: SDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
-                       (ins FR64:$src), "movdq2q\t{$src, $dst|$dst, $src}", [],
-                       IIC_MMX_MOVQ_RR>;
+def MMX_MOVFR642Qrr: MMXSDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
+                              (ins FR64:$src), "movdq2q\t{$src, $dst|$dst, $src}",
+                              [], IIC_MMX_MOVQ_RR>;
 
 def MMX_MOVNTQmr  : MMXI<0xE7, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src),
                          "movntq\t{$src, $dst|$dst, $src}",
@@ -577,6 +576,7 @@ def MMX_MASKMOVQ64: MMXI64<0xF7, MRMSrcReg, (outs), (ins VR64:$src, VR64:$mask),
                            IIC_MMX_MASKMOV>;
 
 // 64-bit bit convert.
+let Predicates = [HasSSE2] in {
 def : Pat<(x86mmx (bitconvert (i64 GR64:$src))),
           (MMX_MOVD64to64rr GR64:$src)>;
 def : Pat<(i64 (bitconvert (x86mmx VR64:$src))),
@@ -585,5 +585,6 @@ def : Pat<(f64 (bitconvert (x86mmx VR64:$src))),
           (MMX_MOVQ2FR64rr VR64:$src)>;
 def : Pat<(x86mmx (bitconvert (f64 FR64:$src))),
           (MMX_MOVFR642Qrr FR64:$src)>;
+}
 
 
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index c834b90..b46da95 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -362,7 +362,7 @@ let Predicates = [HasAVX] in {
   def : Pat<(v16i16 (bitconvert (v32i8 VR256:$src))),  (v16i16 VR256:$src)>;
 }
 
-// Alias instructions that map fld0 to pxor for sse.
+// Alias instructions that map fld0 to xorps for sse or vxorps for avx.
 // This is expanded by ExpandPostRAPseudos.
 let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
     isPseudo = 1 in {
@@ -693,7 +693,7 @@ let Predicates = [HasAVX] in {
             (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
 }
 
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
   let AddedComplexity = 15 in {
   // Move scalar to XMM zero-extended, zeroing a VR128 then do a
   // MOVSS to the lower bits.
@@ -727,7 +727,7 @@ let Predicates = [HasSSE1] in {
             (MOVSSrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR32))>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   let AddedComplexity = 15 in {
   // Move scalar to XMM zero-extended, zeroing a VR128 then do a
   // MOVSD to the lower bits.
@@ -964,10 +964,10 @@ let Predicates = [HasAVX] in {
             (VMOVUPDmr addr:$dst, VR128:$src)>;
 }
 
-let Predicates = [HasSSE1] in
+let Predicates = [UseSSE1] in
   def : Pat<(int_x86_sse_storeu_ps addr:$dst, VR128:$src),
             (MOVUPSmr addr:$dst, VR128:$src)>;
-let Predicates = [HasSSE2] in
+let Predicates = [UseSSE2] in
   def : Pat<(int_x86_sse2_storeu_pd addr:$dst, VR128:$src),
             (MOVUPDmr addr:$dst, VR128:$src)>;
 
@@ -1022,7 +1022,7 @@ let Predicates = [HasAVX] in {
 // Use movaps / movups for SSE integer load / store (one byte shorter).
 // The instructions selected below are then converted to MOVDQA/MOVDQU
 // during the SSE domain pass.
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
   def : Pat<(alignedloadv2i64 addr:$src),
             (MOVAPSrm addr:$src)>;
   def : Pat<(loadv2i64 addr:$src),
@@ -1169,7 +1169,7 @@ let Predicates = [HasAVX] in {
             (VMOVLPDmr addr:$src1, VR128:$src2)>;
 }
 
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
   // (store (vector_shuffle (load addr), v2, <4, 5, 2, 3>), addr) using MOVLPS
   def : Pat<(store (i64 (vector_extract (bc_v2i64 (v4f32 VR128:$src2)),
                                  (iPTR 0))), addr:$src1),
@@ -1194,7 +1194,7 @@ let Predicates = [HasSSE1] in {
             (MOVLPSmr addr:$src1, VR128:$src2)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   // Shuffle with MOVLPD
   def : Pat<(v2f64 (X86Movlpd VR128:$src1, (load addr:$src2))),
             (MOVLPDrm VR128:$src1, addr:$src2)>;
@@ -1268,7 +1268,7 @@ let Predicates = [HasAVX] in {
             (VMOVHPDrm VR128:$src1, addr:$src2)>;
 }
 
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
   // MOVHPS patterns
   def : Pat<(X86Movlhps VR128:$src1,
                  (bc_v4f32 (v2i64 (scalar_to_vector (loadi64 addr:$src2))))),
@@ -1278,7 +1278,7 @@ let Predicates = [HasSSE1] in {
             (MOVHPSrm VR128:$src1, addr:$src2)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   // FIXME: Instead of X86Unpckl, there should be a X86Movlhpd here, the problem
   // is during lowering, where it's not possible to recognize the load fold 
   // cause it has two uses through a bitcast. One use disappears at isel time
@@ -1335,7 +1335,7 @@ let Predicates = [HasAVX] in {
             (VMOVHLPSrr VR128:$src1, VR128:$src2)>;
 }
 
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
   // MOVLHPS patterns
   def : Pat<(v4i32 (X86Movlhps VR128:$src1, VR128:$src2)),
             (MOVLHPSrr VR128:$src1, VR128:$src2)>;
@@ -1622,7 +1622,7 @@ defm VCVTDQ2PSY  : sse12_cvt_p<0x5B, VR256, VR256, i256mem,
 defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, i128mem,
                             "cvtdq2ps\t{$src, $dst|$dst, $src}",
                             SSEPackedSingle, SSE_CVT_PS>,
-                            TB, Requires<[HasSSE2]>;
+                            TB, Requires<[UseSSE2]>;
 
 /// SSE 2 Only
 
@@ -1652,7 +1652,7 @@ def CVTSD2SSrm  : I<0x5A, MRMSrcMem, (outs FR32:$dst), (ins f64mem:$src),
                       [(set FR32:$dst, (fround (loadf64 addr:$src)))],
                       IIC_SSE_CVT_Scalar_RM>,
                       XD,
-                  Requires<[HasSSE2, OptForSize]>;
+                  Requires<[UseSSE2, OptForSize]>;
 
 def Int_VCVTSD2SSrr: I<0x5A, MRMSrcReg,
                        (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
@@ -1673,13 +1673,13 @@ def Int_CVTSD2SSrr: I<0x5A, MRMSrcReg,
                        "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                        [(set VR128:$dst,
                          (int_x86_sse2_cvtsd2ss VR128:$src1, VR128:$src2))],
-                       IIC_SSE_CVT_Scalar_RR>, XD, Requires<[HasSSE2]>;
+                       IIC_SSE_CVT_Scalar_RR>, XD, Requires<[UseSSE2]>;
 def Int_CVTSD2SSrm: I<0x5A, MRMSrcReg,
                        (outs VR128:$dst), (ins VR128:$src1, sdmem:$src2),
                        "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                        [(set VR128:$dst, (int_x86_sse2_cvtsd2ss
                                           VR128:$src1, sse_load_f64:$src2))],
-                       IIC_SSE_CVT_Scalar_RM>, XD, Requires<[HasSSE2]>;
+                       IIC_SSE_CVT_Scalar_RM>, XD, Requires<[UseSSE2]>;
 }
 
 // Convert scalar single to scalar double
@@ -1716,12 +1716,12 @@ def CVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst), (ins FR32:$src),
                    "cvtss2sd\t{$src, $dst|$dst, $src}",
                    [(set FR64:$dst, (fextend FR32:$src))],
                    IIC_SSE_CVT_Scalar_RR>, XS,
-                 Requires<[HasSSE2]>;
+                 Requires<[UseSSE2]>;
 def CVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins f32mem:$src),
                    "cvtss2sd\t{$src, $dst|$dst, $src}",
                    [(set FR64:$dst, (extloadf32 addr:$src))],
                    IIC_SSE_CVT_Scalar_RM>, XS,
-                 Requires<[HasSSE2, OptForSize]>;
+                 Requires<[UseSSE2, OptForSize]>;
 
 // extload f32 -> f64.  This matches load+fextend because we have a hack in
 // the isel (PreprocessForFPConvert) that can introduce loads after dag
@@ -1729,9 +1729,9 @@ def CVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins f32mem:$src),
 // Since these loads aren't folded into the fextend, we have to match it
 // explicitly here.
 def : Pat<(fextend (loadf32 addr:$src)),
-          (CVTSS2SDrm addr:$src)>, Requires<[HasSSE2]>;
+          (CVTSS2SDrm addr:$src)>, Requires<[UseSSE2]>;
 def : Pat<(extloadf32 addr:$src),
-          (CVTSS2SDrr (MOVSSrm addr:$src))>, Requires<[HasSSE2, OptForSpeed]>;
+          (CVTSS2SDrr (MOVSSrm addr:$src))>, Requires<[UseSSE2, OptForSpeed]>;
 
 def Int_VCVTSS2SDrr: I<0x5A, MRMSrcReg,
                       (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
@@ -1751,13 +1751,13 @@ def Int_CVTSS2SDrr: I<0x5A, MRMSrcReg,
                     "cvtss2sd\t{$src2, $dst|$dst, $src2}",
                     [(set VR128:$dst,
                       (int_x86_sse2_cvtss2sd VR128:$src1, VR128:$src2))],
-                    IIC_SSE_CVT_Scalar_RR>, XS, Requires<[HasSSE2]>;
+                    IIC_SSE_CVT_Scalar_RR>, XS, Requires<[UseSSE2]>;
 def Int_CVTSS2SDrm: I<0x5A, MRMSrcMem,
                       (outs VR128:$dst), (ins VR128:$src1, ssmem:$src2),
                     "cvtss2sd\t{$src2, $dst|$dst, $src2}",
                     [(set VR128:$dst,
                       (int_x86_sse2_cvtss2sd VR128:$src1, sse_load_f32:$src2))],
-                    IIC_SSE_CVT_Scalar_RM>, XS, Requires<[HasSSE2]>;
+                    IIC_SSE_CVT_Scalar_RM>, XS, Requires<[UseSSE2]>;
 }
 
 // Convert packed single/double fp to doubleword
@@ -1893,7 +1893,7 @@ let Predicates = [HasAVX] in {
             (VCVTTPS2DQYrm addr:$src)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
             (CVTDQ2PSrr VR128:$src)>;
   def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
@@ -1983,7 +1983,7 @@ def VCVTPS2PDYrm : I<0x5A, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
                      IIC_SSE_CVT_PD_RM>, TB, VEX;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
 def CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        "cvtps2pd\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
@@ -2098,7 +2098,7 @@ let Predicates = [HasAVX] in {
             (VCVTPS2PDYrm addr:$src)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   // Match fextend for 128 conversions
   def : Pat<(v2f64 (X86vfpext (v4f32 VR128:$src))),
             (CVTPS2PDrr VR128:$src)>;
@@ -2325,14 +2325,14 @@ def : Pat<(v4i64 (X86cmpp (v4f64 VR256:$src1), (memop addr:$src2), imm:$cc)),
           (VCMPPDYrmi VR256:$src1, addr:$src2, imm:$cc)>;
 }
 
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
 def : Pat<(v4i32 (X86cmpp (v4f32 VR128:$src1), VR128:$src2, imm:$cc)),
           (CMPPSrri (v4f32 VR128:$src1), (v4f32 VR128:$src2), imm:$cc)>;
 def : Pat<(v4i32 (X86cmpp (v4f32 VR128:$src1), (memop addr:$src2), imm:$cc)),
           (CMPPSrmi (v4f32 VR128:$src1), addr:$src2, imm:$cc)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
 def : Pat<(v2i64 (X86cmpp (v2f64 VR128:$src1), VR128:$src2, imm:$cc)),
           (CMPPDrri VR128:$src1, VR128:$src2, imm:$cc)>;
 def : Pat<(v2i64 (X86cmpp (v2f64 VR128:$src1), (memop addr:$src2), imm:$cc)),
@@ -2409,7 +2409,7 @@ let Predicates = [HasAVX] in {
             (VSHUFPDYrmi VR256:$src1, addr:$src2, imm:$imm)>;
 }
 
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
   def : Pat<(v4i32 (X86Shufp VR128:$src1,
                        (bc_v4i32 (memopv2i64 addr:$src2)), (i8 imm:$imm))),
             (SHUFPSrmi VR128:$src1, addr:$src2, imm:$imm)>;
@@ -2417,7 +2417,7 @@ let Predicates = [HasSSE1] in {
             (SHUFPSrri VR128:$src1, VR128:$src2, imm:$imm)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   // Generic SHUFPD patterns
   def : Pat<(v2i64 (X86Shufp VR128:$src1,
                        (memopv2i64 addr:$src2), (i8 imm:$imm))),
@@ -2518,7 +2518,7 @@ let Predicates = [HasAVX], AddedComplexity = 1 in {
             (VUNPCKLPDrr VR128:$src, VR128:$src)>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   // FIXME: Instead of X86Movddup, there should be a X86Unpckl here, the
   // problem is during lowering, where it's not possible to recognize the load
   // fold cause it has two uses through a bitcast. One use disappears at isel
@@ -2587,16 +2587,16 @@ defm MOVMSKPD : sse12_extr_sign_mask<VR128, int_x86_sse2_movmsk_pd, "movmskpd",
 
 def : Pat<(i32 (X86fgetsign FR32:$src)),
           (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>,
-      Requires<[HasSSE1]>;
+      Requires<[UseSSE1]>;
 def : Pat<(i64 (X86fgetsign FR32:$src)),
           (MOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>,
-      Requires<[HasSSE1]>;
+      Requires<[UseSSE1]>;
 def : Pat<(i32 (X86fgetsign FR64:$src)),
           (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>,
-      Requires<[HasSSE2]>;
+      Requires<[UseSSE2]>;
 def : Pat<(i64 (X86fgetsign FR64:$src)),
           (MOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>,
-      Requires<[HasSSE2]>;
+      Requires<[UseSSE2]>;
 
 //===---------------------------------------------------------------------===//
 // SSE2 - Packed Integer Logical Instructions
@@ -2981,7 +2981,7 @@ multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr,
   def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
                 !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
                 [(set FR32:$dst, (OpNode (load addr:$src)))], itins.rm>, XS,
-            Requires<[HasSSE1, OptForSize]>;
+            Requires<[UseSSE1, OptForSize]>;
   def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                     !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
                     [(set VR128:$dst, (F32Int VR128:$src))], itins.rr>;
@@ -3066,7 +3066,7 @@ multiclass sse2_fp_unop_s<bits<8> opc, string OpcodeStr,
   def SDm : I<opc, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src),
                 !strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
                 [(set FR64:$dst, (OpNode (load addr:$src)))], itins.rm>, XD,
-            Requires<[HasSSE2, OptForSize]>;
+            Requires<[UseSSE2, OptForSize]>;
   def SDr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                     !strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
                     [(set VR128:$dst, (F64Int VR128:$src))], itins.rr>;
@@ -3326,7 +3326,7 @@ def MOVNTDQmr : PDI<0xE7, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
                     IIC_SSE_MOVNT>;
 
 def : Pat<(alignednontemporalstore (v2i64 VR128:$src), addr:$dst),
-          (MOVNTDQmr addr:$dst, VR128:$src)>, Requires<[HasSSE2]>;
+          (MOVNTDQmr addr:$dst, VR128:$src)>, Requires<[UseSSE2]>;
 
 // There is no AVX form for instructions below this point
 def MOVNTImr : I<0xC3, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src),
@@ -3486,7 +3486,7 @@ def MOVDQArr : PDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
 
 def MOVDQUrr :   I<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                    "movdqu\t{$src, $dst|$dst, $src}",
-                   [], IIC_SSE_MOVU_P_RR>, XS, Requires<[HasSSE2]>;
+                   [], IIC_SSE_MOVU_P_RR>, XS, Requires<[UseSSE2]>;
 
 // For Disassembler
 let isCodeGenOnly = 1 in {
@@ -3496,7 +3496,7 @@ def MOVDQArr_REV : PDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
 
 def MOVDQUrr_REV :   I<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
                        "movdqu\t{$src, $dst|$dst, $src}",
-                       [], IIC_SSE_MOVU_P_RR>, XS, Requires<[HasSSE2]>;
+                       [], IIC_SSE_MOVU_P_RR>, XS, Requires<[UseSSE2]>;
 }
 
 let canFoldAsLoad = 1, mayLoad = 1 in {
@@ -3508,7 +3508,7 @@ def MOVDQUrm :   I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
                    "movdqu\t{$src, $dst|$dst, $src}",
                    [/*(set VR128:$dst, (loadv2i64 addr:$src))*/],
                    IIC_SSE_MOVU_P_RM>,
-                 XS, Requires<[HasSSE2]>;
+                 XS, Requires<[UseSSE2]>;
 }
 
 let mayStore = 1 in {
@@ -3520,7 +3520,7 @@ def MOVDQUmr :   I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
                    "movdqu\t{$src, $dst|$dst, $src}",
                    [/*(store (v2i64 VR128:$src), addr:$dst)*/],
                    IIC_SSE_MOVU_P_MR>,
-                 XS, Requires<[HasSSE2]>;
+                 XS, Requires<[UseSSE2]>;
 }
 
 // Intrinsic forms of MOVDQU load and store
@@ -3534,7 +3534,7 @@ def MOVDQUmr_Int :   I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
                        "movdqu\t{$src, $dst|$dst, $src}",
                        [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)],
                        IIC_SSE_MOVU_P_MR>,
-                     XS, Requires<[HasSSE2]>;
+                     XS, Requires<[UseSSE2]>;
 
 } // ExeDomain = SSEPackedInt
 
@@ -4032,7 +4032,7 @@ let Predicates = [HasAVX2] in {
             (VPSRLDQYri VR256:$src1, (BYTE_imm imm:$src2))>;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
   def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2),
             (PSLLDQri VR128:$src1, (BYTE_imm imm:$src2))>;
   def : Pat<(int_x86_sse2_psrl_dq VR128:$src1, imm:$src2),
@@ -4214,7 +4214,7 @@ let Predicates = [HasAVX2] in {
   defm VPSHUFLW : sse2_pshuffle_y<"vpshuflw", v16i16, X86PShuflw>, XD, VEX;
 }
 
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
  let AddedComplexity = 5 in
   defm PSHUFD : sse2_pshuffle<"pshufd", v4i32, X86PShufd>, TB, OpSize;
 
@@ -4377,7 +4377,7 @@ let Predicates = [HasAVX] in {
 }
 
 let Constraints = "$src1 = $dst" in
-  defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[HasSSE2]>;
+  defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[UseSSE2]>;
 
 } // ExeDomain = SSEPackedInt
 
@@ -4661,7 +4661,7 @@ let Predicates = [HasAVX] in {
             (SUBREG_TO_REG (i64 0), (VMOVZQI2PQIrr GR64:$src), sub_xmm)>;
 }
 
-let Predicates = [HasSSE2], AddedComplexity = 20 in {
+let Predicates = [UseSSE2], AddedComplexity = 20 in {
   def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),
             (MOVZDI2PDIrm addr:$src)>;
   def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv2i64 addr:$src)))),
@@ -4701,7 +4701,7 @@ def MOVQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
                     [(set VR128:$dst,
                       (v2i64 (scalar_to_vector (loadi64 addr:$src))))],
                       IIC_SSE_MOVDQ>, XS,
-                    Requires<[HasSSE2]>; // SSE2 instruction with XS Prefix
+                    Requires<[UseSSE2]>; // SSE2 instruction with XS Prefix
 
 //===---------------------------------------------------------------------===//
 // Move Packed Quadword Int to Quadword Int
@@ -4744,7 +4744,7 @@ def MOVZQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
                        (v2i64 (X86vzmovl (v2i64 (scalar_to_vector
                                                  (loadi64 addr:$src))))))],
                                                  IIC_SSE_MOVDQ>,
-                     XS, Requires<[HasSSE2]>;
+                     XS, Requires<[UseSSE2]>;
 
 let Predicates = [HasAVX], AddedComplexity = 20 in {
   def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
@@ -4755,7 +4755,7 @@ let Predicates = [HasAVX], AddedComplexity = 20 in {
             (VMOVZQI2PQIrm addr:$src)>;
 }
 
-let Predicates = [HasSSE2], AddedComplexity = 20 in {
+let Predicates = [UseSSE2], AddedComplexity = 20 in {
   def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
             (MOVZQI2PQIrm addr:$src)>;
   def : Pat<(v2i64 (X86vzmovl (bc_v2i64 (loadv4f32 addr:$src)))),
@@ -4785,7 +4785,7 @@ def MOVZPQILo2PQIrr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                         "movq\t{$src, $dst|$dst, $src}",
                     [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))],
                     IIC_SSE_MOVQ_RR>,
-                      XS, Requires<[HasSSE2]>;
+                      XS, Requires<[UseSSE2]>;
 
 let AddedComplexity = 20 in
 def VMOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
@@ -4800,7 +4800,7 @@ def MOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
                     [(set VR128:$dst, (v2i64 (X86vzmovl
                                              (loadv2i64 addr:$src))))],
                                              IIC_SSE_MOVDQ>,
-                      XS, Requires<[HasSSE2]>;
+                      XS, Requires<[UseSSE2]>;
 }
 
 let AddedComplexity = 20 in {
@@ -4810,7 +4810,7 @@ let AddedComplexity = 20 in {
     def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
               (VMOVZPQILo2PQIrr VR128:$src)>;
   }
-  let Predicates = [HasSSE2] in {
+  let Predicates = [UseSSE2] in {
     def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
               (MOVZPQILo2PQIrm addr:$src)>;
     def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
@@ -4890,7 +4890,7 @@ let Predicates = [HasAVX] in {
             (VMOVSLDUPYrm addr:$src)>;
 }
 
-let Predicates = [HasSSE3] in {
+let Predicates = [UseSSE3] in {
   def : Pat<(v4i32 (X86Movshdup VR128:$src)),
             (MOVSHDUPrr VR128:$src)>;
   def : Pat<(v4i32 (X86Movshdup (bc_v4i32 (memopv2i64 addr:$src)))),
@@ -4959,7 +4959,7 @@ let Predicates = [HasAVX] in {
             (VMOVDDUPYrr VR256:$src)>;
 }
 
-let Predicates = [HasSSE3] in {
+let Predicates = [UseSSE3] in {
   def : Pat<(X86Movddup (memopv2f64 addr:$src)),
             (MOVDDUPrm addr:$src)>;
   def : Pat<(X86Movddup (bc_v2f64 (memopv4f32 addr:$src))),
@@ -5023,7 +5023,7 @@ let Predicates = [HasAVX] in {
                                  f256mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V;
   }
 }
-let Constraints = "$src1 = $dst", Predicates = [HasSSE3] in {
+let Constraints = "$src1 = $dst", Predicates = [UseSSE3] in {
   let ExeDomain = SSEPackedSingle in
   defm ADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "addsubps", VR128,
                               f128mem, SSE_ALU_F32P>, TB, XD;
@@ -5406,7 +5406,7 @@ let Predicates = [HasAVX] in
   defm VPALIGN : ssse3_palign<"vpalignr", 0>, VEX_4V;
 let Predicates = [HasAVX2] in
   defm VPALIGN : ssse3_palign_y<"vpalignr", 0>, VEX_4V;
-let Constraints = "$src1 = $dst", Predicates = [HasSSSE3] in
+let Constraints = "$src1 = $dst", Predicates = [UseSSSE3] in
   defm PALIGN : ssse3_palign<"palignr">;
 
 let Predicates = [HasAVX2] in {
@@ -5431,7 +5431,7 @@ def : Pat<(v16i8 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
           (VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
 }
 
-let Predicates = [HasSSSE3] in {
+let Predicates = [UseSSSE3] in {
 def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
           (PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
 def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
@@ -5565,7 +5565,7 @@ let Predicates = [HasAVX] in {
             (VPMOVZXDQrm addr:$src)>;
 }
 
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
   // Common patterns involving scalar load.
   def : Pat<(int_x86_sse41_pmovsxbw (vzmovl_v2i64 addr:$src)),
             (PMOVSXBWrm addr:$src)>;
@@ -5615,7 +5615,7 @@ let Predicates = [HasAVX] in {
   def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
 }
 
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
   def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
   def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
 }
@@ -5686,7 +5686,7 @@ let Predicates = [HasAVX] in {
             (VPMOVZXWQrm addr:$src)>;
 }
 
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
   // Common patterns involving scalar load
   def : Pat<(int_x86_sse41_pmovsxbd (vzmovl_v4i32 addr:$src)),
             (PMOVSXBDrm addr:$src)>;
@@ -5754,7 +5754,7 @@ let Predicates = [HasAVX] in {
             (VPMOVZXBQrm addr:$src)>;
 }
 
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
   // Common patterns involving scalar load
   def : Pat<(int_x86_sse41_pmovsxbq
               (bitconvert (v4i32 (X86vzmovl
@@ -5900,7 +5900,7 @@ def : Pat<(store (f32 (bitconvert (extractelt (bc_v4i32 (v4f32 VR128:$src1)),
                                               imm:$src2))),
                  addr:$dst),
           (EXTRACTPSmr addr:$dst, VR128:$src1, imm:$src2)>,
-          Requires<[HasSSE41]>;
+          Requires<[UseSSE41]>;
 
 //===----------------------------------------------------------------------===//
 // SSE4.1 - Insert Instructions
@@ -6687,7 +6687,7 @@ def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
 def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
                 (PBLENDVBrm0 VR128:$dst, i128mem:$src2)>;
 
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
   def : Pat<(v16i8 (vselect (v16i8 XMM0), (v16i8 VR128:$src1),
                             (v16i8 VR128:$src2))),
             (PBLENDVBrr0 VR128:$src2, VR128:$src1)>;
@@ -6784,9 +6784,8 @@ multiclass pseudo_pcmpistrm<string asm> {
 }
 
 let Defs = [EFLAGS], usesCustomInserter = 1 in {
-  let AddedComplexity = 1 in
-    defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
-  defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[HasSSE42]>;
+  defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
+  defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[UseSSE42]>;
 }
 
 let Defs = [XMM0, EFLAGS], neverHasSideEffects = 1, Predicates = [HasAVX] in {
@@ -6822,9 +6821,8 @@ multiclass pseudo_pcmpestrm<string asm> {
 }
 
 let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
-  let AddedComplexity = 1 in
-    defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
-  defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[HasSSE42]>;
+  defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
+  defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[UseSSE42]>;
 }
 
 let Predicates = [HasAVX],
diff --git a/lib/Target/X86/X86Subtarget.h b/lib/Target/X86/X86Subtarget.h
index 8d3797e..33608bb 100644
--- a/lib/Target/X86/X86Subtarget.h
+++ b/lib/Target/X86/X86Subtarget.h
@@ -198,6 +198,7 @@ public:
   bool hasSSE42() const { return X86SSELevel >= SSE42; }
   bool hasAVX() const { return X86SSELevel >= AVX; }
   bool hasAVX2() const { return X86SSELevel >= AVX2; }
+  bool hasNoAVX() const { return X86SSELevel < AVX; }
   bool hasSSE4A() const { return HasSSE4A; }
   bool has3DNow() const { return X863DNowLevel >= ThreeDNow; }
   bool has3DNowA() const { return X863DNowLevel >= ThreeDNowA; }
diff --git a/test/CodeGen/X86/fast-isel-x86-64.ll b/test/CodeGen/X86/fast-isel-x86-64.ll
index d8f4663..85a70aa 100644
--- a/test/CodeGen/X86/fast-isel-x86-64.ll
+++ b/test/CodeGen/X86/fast-isel-x86-64.ll
@@ -1,4 +1,5 @@
-; RUN: llc < %s  -fast-isel -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort | FileCheck %s
+; RUN: llc < %s -mattr=-avx -fast-isel -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort | FileCheck %s
+; RUN: llc < %s -mattr=+avx -fast-isel -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort | FileCheck %s --check-prefix=AVX
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
 target triple = "x86_64-apple-darwin10.0.0"
@@ -197,6 +198,11 @@ block2:
 ; CHECK: cvtsi2sdq {{.*}} %xmm0
 ; CHECK: movb $1, %al
 ; CHECK: callq _test16callee
+
+; AVX: movabsq $1
+; AVX: vmovsd LCP{{.*}}_{{.*}}(%rip), %xmm0
+; AVX: movb $1, %al
+; AVX: callq _test16callee
   call void (...)* @test16callee(double 1.000000e+00)
   ret void
 }
diff --git a/test/CodeGen/X86/widen_load-1.ll b/test/CodeGen/X86/widen_load-1.ll
index 9705d14..dfaa3d6 100644
--- a/test/CodeGen/X86/widen_load-1.ll
+++ b/test/CodeGen/X86/widen_load-1.ll
@@ -1,12 +1,17 @@
-; RUN: llc %s -o - -march=x86-64 -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
+; RUN: llc %s -o - -march=x86-64 -mattr=-avx -mtriple=x86_64-unknown-linux-gnu | FileCheck %s --check-prefix=SSE
+; RUN: llc %s -o - -march=x86-64 -mattr=+avx -mtriple=x86_64-unknown-linux-gnu | FileCheck %s --check-prefix=AVX
 ; PR4891
 ; PR5626
 
 ; This load should be before the call, not after.
 
-; CHECK: movaps    compl+128(%rip), %xmm0
-; CHECK: movaps  %xmm0, (%rsp)
-; CHECK: callq   killcommon
+; SSE: movaps    compl+128(%rip), %xmm0
+; SSE: movaps  %xmm0, (%rsp)
+; SSE: callq   killcommon
+
+; AVX: vmovapd    compl+128(%rip), %xmm0
+; AVX: vmovapd  %xmm0, (%rsp)
+; AVX: callq   killcommon
 
 @compl = linkonce global [20 x i64] zeroinitializer, align 64 ; <[20 x i64]*> [#uses=1]