[X86] Simplify some of the type checks in combineSubToSubus.

If we have SSE2 we can handle any i8/i16 type and let
type legalization deal with it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@368538 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 7c82d6d126d5ec5f3657f5e2708acd927762ce87..4bbb69fce5a6152e8a6337963022b7aaba95f7f2 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -42860,13 +42860,15 @@ static SDValue combineSubToSubus(SDNode *N, SelectionDAG &DAG,
   SDValue Op1 = N->getOperand(1);
   EVT VT = N->getValueType(0);
 
+  if (!VT.isVector())
+    return SDValue();
+
   // PSUBUS is supported, starting from SSE2, but truncation for v8i32
   // is only worth it with SSSE3 (PSHUFB).
-  if (!(Subtarget.hasSSE2() && (VT == MVT::v16i8 || VT == MVT::v8i16)) &&
+  EVT EltVT = VT.getVectorElementType();
+  if (!(Subtarget.hasSSE2() && (EltVT == MVT::i8 || EltVT == MVT::i16)) &&
       !(Subtarget.hasSSSE3() && (VT == MVT::v8i32 || VT == MVT::v8i64)) &&
-      !(Subtarget.hasAVX() && (VT == MVT::v32i8 || VT == MVT::v16i16)) &&
-      !(Subtarget.useBWIRegs() && (VT == MVT::v64i8 || VT == MVT::v32i16 ||
-                                   VT == MVT::v16i32 || VT == MVT::v8i64)))
+      !(Subtarget.useBWIRegs() && (VT == MVT::v16i32)))
     return SDValue();
 
   SDValue SubusLHS, SubusRHS;
@@ -42898,9 +42900,12 @@ static SDValue combineSubToSubus(SDNode *N, SelectionDAG &DAG,
 
   // PSUBUS doesn't support v8i32/v8i64/v16i32, but it can be enabled with
   // special preprocessing in some cases.
-  if (VT != MVT::v8i32 && VT != MVT::v16i32 && VT != MVT::v8i64)
+  if (EltVT == MVT::i8 || EltVT == MVT::i16)
     return DAG.getNode(ISD::USUBSAT, SDLoc(N), VT, SubusLHS, SubusRHS);
 
+  assert((VT == MVT::v8i32 || VT == MVT::v16i32 || VT == MVT::v8i64) &&
+         "Unexpected VT!");
+
   // Special preprocessing case can be only applied
   // if the value was zero extended from 16 bit,
   // so we require first 16 bits to be zeros for 32 bit