[EraVM] Fix xor optimization bug

This issue occurs when DAGCombiner in visitXOR incorrectly folds !(x cc y) to (x !cc y) because isConstTrueVal returns true for -1 constant. Fix this issue by setting that true and false are represented with 1 and 0 values. Signed-off-by: Vladimir Radosavljevic <[email protected]>
matter-labs · Apr 2, 2024 · 51fc815 · 51fc815
1 parent 122a2f8
commit 51fc815
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 1 deletion.
diff --git a/llvm/lib/Target/EraVM/EraVMISelLowering.cpp b/llvm/lib/Target/EraVM/EraVMISelLowering.cpp
@@ -105,6 +105,8 @@ EraVMTargetLowering::EraVMTargetLowering(const TargetMachine &TM,
   // Provide all sorts of operation actions
   setStackPointerRegisterToSaveRestore(EraVM::SP);
 
+  setBooleanContents(ZeroOrOneBooleanContent);
+
   // By default, expand all i256bit operations
   for (unsigned Opc = 0; Opc < ISD::BUILTIN_OP_END; ++Opc)
     setOperationAction(Opc, MVT::i256, Expand);

diff --git a/llvm/test/CodeGen/EraVM/xor-opt-bug.ll b/llvm/test/CodeGen/EraVM/xor-opt-bug.ll
@@ -9,7 +9,9 @@ define i256 @test(i256 %a) {
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    sub! r1, r0, r1
 ; CHECK-NEXT:    add 0, r0, r1
-; CHECK-NEXT:    add.eq 1, r0, r1
+; CHECK-NEXT:    add.ne 1, r0, r1
+; CHECK-NEXT:    sub.s 1, r0, r2
+; CHECK-NEXT:    xor r1, r2, r1
 ; CHECK-NEXT:    ret
   %cmp = icmp ne i256 %a, 0
   %zext = zext i1 %cmp to i256