[llvm] r357029 - [SDAG] add simplifications for FP at node creation time

[Sanjay Patel via llvm-commits]

Author: spatel
Date: Tue Mar 26 13:54:15 2019
New Revision: 357029

URL: http://llvm.org/viewvc/llvm-project?rev=357029&view=rev
Log:
[SDAG] add simplifications for FP at node creation time

We have the folds for fadd/fsub/fmul already in DAGCombiner,
so it may be possible to remove that code if we can guarantee that
these ops are zapped before they can exist.

Modified:
    llvm/trunk/include/llvm/CodeGen/SelectionDAG.h
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
    llvm/trunk/test/CodeGen/X86/extract-fp.ll

Modified: llvm/trunk/include/llvm/CodeGen/SelectionDAG.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/SelectionDAG.h?rev=357029&r1=357028&r2=357029&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/SelectionDAG.h (original)
+++ llvm/trunk/include/llvm/CodeGen/SelectionDAG.h Tue Mar 26 13:54:15 2019
@@ -976,6 +976,10 @@ public:
   /// Try to simplify a shift into 1 of its operands or a constant.
   SDValue simplifyShift(SDValue X, SDValue Y);
 
+  /// Try to simplify a floating-point binary operation into 1 of its operands
+  /// or a constant.
+  SDValue simplifyFPBinop(unsigned Opcode, SDValue X, SDValue Y);
+
   /// VAArg produces a result and token chain, and takes a pointer
   /// and a source value as input.
   SDValue getVAArg(EVT VT, const SDLoc &dl, SDValue Chain, SDValue Ptr,

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp?rev=357029&r1=357028&r2=357029&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp Tue Mar 26 13:54:15 2019
@@ -4931,6 +4931,8 @@ SDValue SelectionDAG::getNode(unsigned O
     assert(VT.isFloatingPoint() && "This operator only applies to FP types!");
     assert(N1.getValueType() == N2.getValueType() &&
            N1.getValueType() == VT && "Binary operator types must match!");
+    if (SDValue V = simplifyFPBinop(Opcode, N1, N2))
+      return V;
     break;
   case ISD::FCOPYSIGN:   // N1 and result must match.  N1/N2 need not match.
     assert(N1.getValueType() == VT &&
@@ -7052,6 +7054,31 @@ SDValue SelectionDAG::simplifyShift(SDVa
 
   return SDValue();
 }
+
+// TODO: Use fast-math-flags to enable more simplifications.
+SDValue SelectionDAG::simplifyFPBinop(unsigned Opcode, SDValue X, SDValue Y) {
+  ConstantFPSDNode *YC = isConstOrConstSplatFP(Y, /* AllowUndefs */ true);
+  if (!YC)
+    return SDValue();
+
+  // X + -0.0 --> X
+  if (Opcode == ISD::FADD)
+    if (YC->getValueAPF().isNegZero())
+      return X;
+
+  // X - +0.0 --> X
+  if (Opcode == ISD::FSUB)
+    if (YC->getValueAPF().isPosZero())
+      return X;
+
+  // X * 1.0 --> X
+  // X / 1.0 --> X
+  if (Opcode == ISD::FMUL || Opcode == ISD::FDIV)
+    if (YC->getValueAPF().isExactlyValue(1.0))
+      return X;
+
+  return SDValue();
+}
 
 SDValue SelectionDAG::getVAArg(EVT VT, const SDLoc &dl, SDValue Chain,
                                SDValue Ptr, SDValue SV, unsigned Align) {

Modified: llvm/trunk/test/CodeGen/X86/extract-fp.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/extract-fp.ll?rev=357029&r1=357028&r2=357029&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/extract-fp.ll (original)
+++ llvm/trunk/test/CodeGen/X86/extract-fp.ll Tue Mar 26 13:54:15 2019
@@ -36,12 +36,11 @@ define float @ext_fmul_v4f32(<4 x float>
   ret float %ext
 }
 
-; TODO: X / 1.0 --> X
+; X / 1.0 --> X
 
 define float @ext_fdiv_v4f32(<4 x float> %x) {
 ; CHECK-LABEL: ext_fdiv_v4f32:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    divss {{.*}}(%rip), %xmm0
 ; CHECK-NEXT:    retq
   %bo = fdiv <4 x float> %x, <float 1.0, float 2.0, float 3.0, float 42.0>
   %ext = extractelement <4 x float> %bo, i32 0