[llvm] 2489cba - [PowerPC] Fix infinite loop in peephole CR optimization (PR49509)

[Nikita Popov via llvm-commits]

Author: Nikita Popov
Date: 2021-03-11T14:25:22+01:00
New Revision: 2489cbaa8057c736475fd88990f4f6dbf022873d

URL: https://github.com/llvm/llvm-project/commit/2489cbaa8057c736475fd88990f4f6dbf022873d
DIFF: https://github.com/llvm/llvm-project/commit/2489cbaa8057c736475fd88990f4f6dbf022873d.diff

LOG: [PowerPC] Fix infinite loop in peephole CR optimization (PR49509)

If we encounter a degenerate select node where both operands are
the same, then we can continue negating the condition while swapping
operands, resulting in an infinite loop. Avoid this by bailing out
if both operands are the same.

Fixes https://bugs.llvm.org/show_bug.cgi?id=49509.

Differential Revision: https://reviews.llvm.org/D98340

Added: 
    llvm/test/CodeGen/PowerPC/pr49509.ll

Modified: 
    llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
index a8aa97eca76c..2f341e43de45 100644
--- a/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
@@ -5937,7 +5937,13 @@ bool PPCDAGToDAGISel::AllUsersSelectZero(SDNode *N) {
         User->getMachineOpcode() != PPC::SELECT_I8)
       return false;
 
+    SDNode *Op1 = User->getOperand(1).getNode();
     SDNode *Op2 = User->getOperand(2).getNode();
+    // If we have a degenerate select with two equal operands, swapping will
+    // not do anything, and we may run into an infinite loop.
+    if (Op1 == Op2)
+      return false;
+
     if (!Op2->isMachineOpcode())
       return false;
 

diff  --git a/llvm/test/CodeGen/PowerPC/pr49509.ll b/llvm/test/CodeGen/PowerPC/pr49509.ll
new file mode 100644
index 000000000000..f13733c18047
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/pr49509.ll
@@ -0,0 +1,81 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=powerpc-unknown-linux-gnu < %s | FileCheck %s
+
+target datalayout = "E-m:e-p:32:32-i64:64-n32"
+
+define void @test() {
+; CHECK-LABEL: test:
+; CHECK:       # %bb.0: # %bb
+; CHECK-NEXT:    bc 12, 20, .LBB0_2
+; CHECK-NEXT:  # %bb.1: # %bb2
+; CHECK-NEXT:    li 3, 0
+; CHECK-NEXT:    stw 3, 0(3)
+; CHECK-NEXT:    lis 3, 256
+; CHECK-NEXT:    stw 3, 0(3)
+; CHECK-NEXT:    blr
+; CHECK-NEXT:  .LBB0_2: # %bb1
+; CHECK-NEXT:    bclr 4, 20, 0
+; CHECK-NEXT:  # %bb.3: # %bb66
+; CHECK-NEXT:    lwz 4, 12(0)
+; CHECK-NEXT:    lwz 5, 8(0)
+; CHECK-NEXT:    lwz 6, 0(0)
+; CHECK-NEXT:    lwz 7, 4(0)
+; CHECK-NEXT:    lbz 3, 0(3)
+; CHECK-NEXT:    and 5, 5, 6
+; CHECK-NEXT:    and 4, 4, 7
+; CHECK-NEXT:    and 4, 4, 5
+; CHECK-NEXT:    cmpwi 3, 0
+; CHECK-NEXT:    lis 3, 256
+; CHECK-NEXT:    lis 7, 512
+; CHECK-NEXT:    bc 12, 2, .LBB0_4
+; CHECK-NEXT:    b .LBB0_5
+; CHECK-NEXT:  .LBB0_4: # %bb66
+; CHECK-NEXT:    li 3, 0
+; CHECK-NEXT:  .LBB0_5: # %bb66
+; CHECK-NEXT:    cmpwi 1, 4, -1
+; CHECK-NEXT:    cmpwi 5, 4, -1
+; CHECK-NEXT:    li 6, 0
+; CHECK-NEXT:    bc 12, 6, .LBB0_6
+; CHECK-NEXT:    b .LBB0_7
+; CHECK-NEXT:  .LBB0_6: # %bb66
+; CHECK-NEXT:    addi 3, 7, 0
+; CHECK-NEXT:  .LBB0_7: # %bb66
+; CHECK-NEXT:    cror 20, 22, 2
+; CHECK-NEXT:    stw 3, 0(3)
+; CHECK-NEXT:    bc 12, 20, .LBB0_9
+; CHECK-NEXT:  # %bb.8: # %bb66
+; CHECK-NEXT:    ori 3, 6, 0
+; CHECK-NEXT:    b .LBB0_10
+; CHECK-NEXT:  .LBB0_9: # %bb66
+; CHECK-NEXT:    li 3, 0
+; CHECK-NEXT:  .LBB0_10: # %bb66
+; CHECK-NEXT:    stw 3, 0(3)
+; CHECK-NEXT:    blr
+bb:
+  br i1 undef, label %bb2, label %bb1
+
+bb2:                                              ; preds = %bb
+  %i = select i1 undef, i64 0, i64 72057594037927936
+  store i64 %i, i64* undef, align 8
+  ret void
+
+bb1:                                              ; preds = %bb
+  %i50 = load i8, i8* undef, align 8
+  %i52 = load i128, i128* null, align 8
+  %i62 = icmp eq i8 %i50, 0
+  br i1 undef, label %bb66, label %bb64
+
+bb64:                                             ; preds = %bb63
+  ret void
+
+bb66:                                             ; preds = %bb63
+  %i67 = lshr i128 -1, 0
+  %i68 = xor i128 %i52, -1
+  %i69 = add i128 0, %i68
+  %i70 = and i128 %i67, %i69
+  %i71 = icmp eq i128 %i70, 0
+  %i74 = select i1 %i62, i64 0, i64 72057594037927936
+  %i75 = select i1 %i71, i64 144115188075855872, i64 %i74
+  store i64 %i75, i64* undef, align 8
+  ret void
+}