Convert V_SETALLONES/AVX_SETALLONES/AVX2_SETALLONES to Post-RA pseudos.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@162740 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp
index 846e8b934abb82c7999e27846b2343ad2f94f034..8d4a2b4e17b13a17291f813ffd0eb906f1716bee 100644 (file)
--- a/lib/Target/X86/X86InstrInfo.cpp
+++ b/lib/Target/X86/X86InstrInfo.cpp
@@ -3432,6 +3432,10 @@ bool X86InstrInfo::expandPostRAPseudo(MachineBasicBlock::iterator MI) const {
   case X86::AVX_SET0:
     assert(HasAVX && "AVX not supported");
     return Expand2AddrUndef(MI, get(X86::VXORPSYrr));
+  case X86::V_SETALLONES:
+    return Expand2AddrUndef(MI, get(HasAVX ? X86::VPCMPEQDrr : X86::PCMPEQDrr));
+  case X86::AVX2_SETALLONES:
+    return Expand2AddrUndef(MI, get(X86::VPCMPEQDYrr));
   case X86::TEST8ri_NOREX:
     MI->setDesc(get(X86::TEST8ri));
     return true;
@@ -3789,7 +3793,6 @@ MachineInstr* X86InstrInfo::foldMemoryOperandImpl(MachineFunction &MF,
       break;
     case X86::V_SET0:
     case X86::V_SETALLONES:
-    case X86::AVX_SETALLONES:
       Alignment = 16;
       break;
     case X86::FsFLD0SD:
@@ -3825,7 +3828,6 @@ MachineInstr* X86InstrInfo::foldMemoryOperandImpl(MachineFunction &MF,
   switch (LoadMI->getOpcode()) {
   case X86::V_SET0:
   case X86::V_SETALLONES:
-  case X86::AVX_SETALLONES:
   case X86::AVX2_SETALLONES:
   case X86::AVX_SET0:
   case X86::FsFLD0SD:
@@ -3864,8 +3866,7 @@ MachineInstr* X86InstrInfo::foldMemoryOperandImpl(MachineFunction &MF,
     else
       Ty = VectorType::get(Type::getInt32Ty(MF.getFunction()->getContext()), 4);
 
-    bool IsAllOnes = (Opc == X86::V_SETALLONES || Opc == X86::AVX_SETALLONES ||
-                      Opc == X86::AVX2_SETALLONES);
+    bool IsAllOnes = (Opc == X86::V_SETALLONES || Opc == X86::AVX2_SETALLONES);
     const Constant *C = IsAllOnes ? Constant::getAllOnesValue(Ty) :
                                     Constant::getNullValue(Ty);
     unsigned CPI = MCP.getConstantPoolIndex(C, Alignment);

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index a9359ac857573afe252c19f7a9142991e529ed97..52da79aab026f0a8eef8acbce4eaa883fafed2cb 100644 (file)
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -436,19 +436,13 @@ def : Pat<(bc_v4i64 (v8f32 immAllZerosV)),
 
 // We set canFoldAsLoad because this can be converted to a constant-pool
 // load of an all-ones value if folding it would be beneficial.
-// FIXME: Change encoding to pseudo! This is blocked right now by the x86
-// JIT implementation, it does not expand the instructions below like
-// X86MCInstLower does.
 let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
-    isCodeGenOnly = 1, ExeDomain = SSEPackedInt in {
-  let Predicates = [HasAVX] in
-  def AVX_SETALLONES : PDI<0x76, MRMInitReg, (outs VR128:$dst), (ins), "",
-                         [(set VR128:$dst, (v4i32 immAllOnesV))]>, VEX_4V;
-  def V_SETALLONES : PDI<0x76, MRMInitReg, (outs VR128:$dst), (ins), "",
-                         [(set VR128:$dst, (v4i32 immAllOnesV))]>;
+    isPseudo = 1 in {
+  def V_SETALLONES : I<0, Pseudo, (outs VR128:$dst), (ins), "",
+                       [(set VR128:$dst, (v4i32 immAllOnesV))]>;
   let Predicates = [HasAVX2] in
-  def AVX2_SETALLONES : PDI<0x76, MRMInitReg, (outs VR256:$dst), (ins), "",
-                          [(set VR256:$dst, (v8i32 immAllOnesV))]>, VEX_4V;
+  def AVX2_SETALLONES : I<0, Pseudo, (outs VR256:$dst), (ins), "",
+                          [(set VR256:$dst, (v8i32 immAllOnesV))]>;
 }
 
 

diff --git a/lib/Target/X86/X86MCInstLower.cpp b/lib/Target/X86/X86MCInstLower.cpp
index 4ee03e5a27ee4888b758e94ef1ac91de5a605d81..1c2ef25aca5fc785a9bc0a5b131470ae58b28814 100644 (file)
--- a/lib/Target/X86/X86MCInstLower.cpp
+++ b/lib/Target/X86/X86MCInstLower.cpp
@@ -377,9 +377,6 @@ ReSimplify:
   case X86::SETB_C64r:    LowerUnaryToTwoAddr(OutMI, X86::SBB64rr); break;
   case X86::MOV8r0:       LowerUnaryToTwoAddr(OutMI, X86::XOR8rr); break;
   case X86::MOV32r0:      LowerUnaryToTwoAddr(OutMI, X86::XOR32rr); break;
-  case X86::V_SETALLONES:  LowerUnaryToTwoAddr(OutMI, X86::PCMPEQDrr); break;
-  case X86::AVX_SETALLONES:  LowerUnaryToTwoAddr(OutMI, X86::VPCMPEQDrr); break;
-  case X86::AVX2_SETALLONES: LowerUnaryToTwoAddr(OutMI, X86::VPCMPEQDYrr);break;
 
   case X86::MOV16r0:
     LowerSubReg32_Op0(OutMI, X86::MOV32r0);   // MOV16r0 -> MOV32r0