[llvm] [DAG] getCarry - always succeed if we encounter a i1 type during trunc/ext peeling (PR #169777)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-selectiondag

Author: Simon Pilgrim (RKSimon)

<details>
<summary>Changes</summary>

If we are force reconstructing a carry from a raw MVT::i1 type, make sure we don't miss any cases while peeling through trunc/ext chains - check for i1 types at the start of the while loop

Fixes #<!-- -->169691

---
Full diff: https://github.com/llvm/llvm-project/pull/169777.diff


2 Files Affected:

- (modified) llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (+3-3) 
- (modified) llvm/test/CodeGen/X86/addcarry.ll (+3-12) 


``````````diff
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 6b79dbb46cadc..0f3a207cc6414 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -3288,6 +3288,9 @@ static SDValue getAsCarry(const TargetLowering &TLI, SDValue V,
 
   // First, peel away TRUNCATE/ZERO_EXTEND/AND nodes due to legalization.
   while (true) {
+    if (ForceCarryReconstruction && V.getValueType() == MVT::i1)
+      return V;
+
     if (V.getOpcode() == ISD::TRUNCATE || V.getOpcode() == ISD::ZERO_EXTEND) {
       V = V.getOperand(0);
       continue;
@@ -3302,9 +3305,6 @@ static SDValue getAsCarry(const TargetLowering &TLI, SDValue V,
       continue;
     }
 
-    if (ForceCarryReconstruction && V.getValueType() == MVT::i1)
-      return V;
-
     break;
   }
 
diff --git a/llvm/test/CodeGen/X86/addcarry.ll b/llvm/test/CodeGen/X86/addcarry.ll
index f8a04f8514988..ee4482062df31 100644
--- a/llvm/test/CodeGen/X86/addcarry.ll
+++ b/llvm/test/CodeGen/X86/addcarry.ll
@@ -1517,18 +1517,9 @@ define i1 @pr84831(i64 %arg) {
 define void @pr169691(ptr %p0, i64 %implicit, i1 zeroext %carry) {
 ; CHECK-LABEL: pr169691:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    movq (%rdi), %rax
-; CHECK-NEXT:    addq %rsi, %rax
-; CHECK-NEXT:    setb %cl
-; CHECK-NEXT:    movl %edx, %edx
-; CHECK-NEXT:    addq %rax, %rdx
-; CHECK-NEXT:    setb %al
-; CHECK-NEXT:    orb %cl, %al
-; CHECK-NEXT:    movq %rdx, (%rdi)
-; CHECK-NEXT:    addq 8(%rdi), %rsi
-; CHECK-NEXT:    movzbl %al, %eax
-; CHECK-NEXT:    addq %rsi, %rax
-; CHECK-NEXT:    movq %rax, 8(%rdi)
+; CHECK-NEXT:    addb $-1, %dl
+; CHECK-NEXT:    adcq %rsi, (%rdi)
+; CHECK-NEXT:    adcq %rsi, 8(%rdi)
 ; CHECK-NEXT:    retq
   %a0 = load i64, ptr %p0, align 8
   %uaddo0 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %a0, i64 %implicit)

``````````

</details>


https://github.com/llvm/llvm-project/pull/169777