2 files changed, 38 insertions, 32 deletions

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 1646c0fd39a..e6f7e623067 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -1208,7 +1208,6 @@ static Value *simplifyMaskedLoad(const IntrinsicInst &II,
 }
 
 // TODO, Obvious Missing Transforms:
-// * SimplifyDemandedVectorElts
 // * Single constant active lane -> store
 // * Narrow width by halfs excluding zero/undef lanes
 Instruction *InstCombiner::simplifyMaskedStore(IntrinsicInst &II) {
@@ -1244,6 +1243,8 @@ Instruction *InstCombiner::simplifyMaskedStore(IntrinsicInst &II) {
 // * Dereferenceable address & few lanes -> scalarize speculative load/selects
 // * Adjacent vector addresses -> masked.load
 // * Narrow width by halfs excluding zero/undef lanes
+// * Vector splat address w/known mask -> scalar load
+// * Vector incrementing address -> vector masked load
 static Instruction *simplifyMaskedGather(IntrinsicInst &II, InstCombiner &IC) {
   // If the mask is all zeros, return the "passthru" argument of the gather.
   auto *ConstMask = dyn_cast<Constant>(II.getArgOperand(2));
@@ -1253,6 +1254,38 @@ static Instruction *simplifyMaskedGather(IntrinsicInst &II, InstCombiner &IC) {
   return nullptr;
 }
 
+// TODO, Obvious Missing Transforms:
+// * Single constant active lane -> store
+// * Adjacent vector addresses -> masked.store
+// * Narrow store width by halfs excluding zero/undef lanes
+// * Vector splat address w/known mask -> scalar store
+// * Vector incrementing address -> vector masked store
+Instruction *InstCombiner::simplifyMaskedScatter(IntrinsicInst &II) {
+  auto *ConstMask = dyn_cast<Constant>(II.getArgOperand(3));
+  if (!ConstMask)
+    return nullptr;
+
+  // If the mask is all zeros, a scatter does nothing.
+  if (ConstMask->isNullValue())
+    return eraseInstFromFunction(II);
+
+  // Use masked off lanes to simplify operands via SimplifyDemandedVectorElts
+  APInt DemandedElts = possiblyDemandedEltsInMask(ConstMask);
+  APInt UndefElts(DemandedElts.getBitWidth(), 0);
+  if (Value *V = SimplifyDemandedVectorElts(II.getOperand(0),
+                                            DemandedElts, UndefElts)) {
+    II.setOperand(0, V);
+    return &II;
+  }
+  if (Value *V = SimplifyDemandedVectorElts(II.getOperand(1),
+                                            DemandedElts, UndefElts)) {
+    II.setOperand(1, V);
+    return &II;
+  }
+
+  return nullptr;
+}
+
 /// This function transforms launder.invariant.group and strip.invariant.group
 /// like:
 /// launder(launder(%x)) -> launder(%x)       (the result is not the argument)
@@ -1287,37 +1320,6 @@ static Instruction *simplifyInvariantGroupIntrinsic(IntrinsicInst &II,
   return cast<Instruction>(Result);
 }
 
-// TODO, Obvious Missing Transforms:
-// * SimplifyDemandedVectorElts
-// * Single constant active lane -> store
-// * Adjacent vector addresses -> masked.store
-// * Narrow store width by halfs excluding zero/undef lanes
-Instruction *InstCombiner::simplifyMaskedScatter(IntrinsicInst &II) {
-  auto *ConstMask = dyn_cast<Constant>(II.getArgOperand(3));
-  if (!ConstMask)
-    return nullptr;
-
-  // If the mask is all zeros, a scatter does nothing.
-  if (ConstMask->isNullValue())
-    return eraseInstFromFunction(II);
-
-  // Use masked off lanes to simplify operands via SimplifyDemandedVectorElts
-  APInt DemandedElts = possiblyDemandedEltsInMask(ConstMask);
-  APInt UndefElts(DemandedElts.getBitWidth(), 0);
-  if (Value *V = SimplifyDemandedVectorElts(II.getOperand(0),
-                                            DemandedElts, UndefElts)) {
-    II.setOperand(0, V);
-    return &II;
-  }
-  if (Value *V = SimplifyDemandedVectorElts(II.getOperand(1),
-                                            DemandedElts, UndefElts)) {
-    II.setOperand(1, V);
-    return &II;
-  }
-
-  return nullptr;
-}
-
 static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombiner &IC) {
   assert((II.getIntrinsicID() == Intrinsic::cttz ||
           II.getIntrinsicID() == Intrinsic::ctlz) &&
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 3462149a002..4cc81b9e50f 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1568,6 +1568,10 @@ Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
         return replaceInstUsesWith(GEP, V);
       return &GEP;
     }
+
+    // TODO: 1) Scalarize splat operands, 2) scalarize entire instruction if
+    // possible (decide on canonical form for pointer broadcast), 3) exploit
+    // undef elements to decrease demanded bits  
   }
 
   Value *PtrOp = GEP.getOperand(0);