[DAGCombiner] Allow tryToFoldExtOfLoad to use a sextload for zext nneg. (#81714)

If the load is used by any signed setccs, we can use a sextload
instead of zextload. Then we don't have to give up on extending
the load.

2 files changed, 57 insertions, 10 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 18037f11ee40..2a09e44e1929 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -13171,20 +13171,39 @@ static SDValue tryToFoldExtOfExtload(SelectionDAG &DAG, DAGCombiner &Combiner,
 
 // fold ([s|z]ext (load x)) -> ([s|z]ext (truncate ([s|z]extload x)))
 // Only generate vector extloads when 1) they're legal, and 2) they are
-// deemed desirable by the target.
+// deemed desirable by the target. NonNegZExt can be set to true if a zero
+// extend has the nonneg flag to allow use of sextload if profitable.
 static SDValue tryToFoldExtOfLoad(SelectionDAG &DAG, DAGCombiner &Combiner,
                                   const TargetLowering &TLI, EVT VT,
                                   bool LegalOperations, SDNode *N, SDValue N0,
                                   ISD::LoadExtType ExtLoadType,
-                                  ISD::NodeType ExtOpc) {
+                                  ISD::NodeType ExtOpc,
+                                  bool NonNegZExt = false) {
+  if (!ISD::isNON_EXTLoad(N0.getNode()) || !ISD::isUNINDEXEDLoad(N0.getNode()))
+    return {};
+
+  // If this is zext nneg, see if it would make sense to treat it as a sext.
+  if (NonNegZExt) {
+    assert(ExtLoadType == ISD::ZEXTLOAD && ExtOpc == ISD::ZERO_EXTEND &&
+           "Unexpected load type or opcode");
+    for (SDNode *User : N0->uses()) {
+      if (User->getOpcode() == ISD::SETCC) {
+        ISD::CondCode CC = cast<CondCodeSDNode>(User->getOperand(2))->get();
+        if (ISD::isSignedIntSetCC(CC)) {
+          ExtLoadType = ISD::SEXTLOAD;
+          ExtOpc = ISD::SIGN_EXTEND;
+          break;
+        }
+      }
+    }
+  }
+
   // TODO: isFixedLengthVector() should be removed and any negative effects on
   // code generation being the result of that target's implementation of
   // isVectorLoadExtDesirable().
-  if (!ISD::isNON_EXTLoad(N0.getNode()) ||
-      !ISD::isUNINDEXEDLoad(N0.getNode()) ||
-      ((LegalOperations || VT.isFixedLengthVector() ||
-        !cast<LoadSDNode>(N0)->isSimple()) &&
-       !TLI.isLoadExtLegal(ExtLoadType, VT, N0.getValueType())))
+  if ((LegalOperations || VT.isFixedLengthVector() ||
+       !cast<LoadSDNode>(N0)->isSimple()) &&
+      !TLI.isLoadExtLegal(ExtLoadType, VT, N0.getValueType()))
     return {};
 
   bool DoXform = true;
@@ -13780,9 +13799,9 @@ SDValue DAGCombiner::visitZERO_EXTEND(SDNode *N) {
   }
 
   // Try to simplify (zext (load x)).
-  if (SDValue foldedExt =
-          tryToFoldExtOfLoad(DAG, *this, TLI, VT, LegalOperations, N, N0,
-                             ISD::ZEXTLOAD, ISD::ZERO_EXTEND))
+  if (SDValue foldedExt = tryToFoldExtOfLoad(
+          DAG, *this, TLI, VT, LegalOperations, N, N0, ISD::ZEXTLOAD,
+          ISD::ZERO_EXTEND, N->getFlags().hasNonNeg()))
     return foldedExt;
 
   if (SDValue foldedExt =
diff --git a/llvm/test/CodeGen/RISCV/load-setcc-combine.ll b/llvm/test/CodeGen/RISCV/load-setcc-combine.ll
new file mode 100644
index 000000000000..403999b93126
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/load-setcc-combine.ll
@@ -0,0 +1,28 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc < %s -mtriple=riscv64 | FileCheck %s
+
+define i8 @zext_nonneg_load_i16(ptr %x, ptr %y) {
+; CHECK-LABEL: zext_nonneg_load_i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    lh a0, 0(a0)
+; CHECK-NEXT:    bltz a0, .LBB0_2
+; CHECK-NEXT:  # %bb.1: # %cont
+; CHECK-NEXT:    add a0, a1, a0
+; CHECK-NEXT:    lbu a0, 0(a0)
+; CHECK-NEXT:    ret
+; CHECK-NEXT:  .LBB0_2: # %exit
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+  %a = load i16, ptr %x
+  %b = icmp slt i16 %a, 0
+  br i1 %b, label %exit, label %cont
+
+cont:
+  %c = zext nneg i16 %a to i64
+  %d = getelementptr i8, ptr %y, i64 %c
+  %e = load i8, ptr %d
+  ret i8 %e
+
+exit:
+  ret i8 0
+}