[llvm] [GlobalISel] Introduce `G_POISON` (PR #127825)

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Tue Apr 22 10:17:29 PDT 2025


Mateusz =?utf-8?q?Sokół?= <mat646 at gmail.com>,
Mateusz =?utf-8?q?Sokół?= <mat646 at gmail.com>,
Mateusz =?utf-8?q?Sokół?= <mat646 at gmail.com>,
Mateusz =?utf-8?q?Sokół?= <mat646 at gmail.com>,
Mateusz =?utf-8?q?Sokół?= <mat646 at gmail.com>,
Mateusz =?utf-8?q?Sokół?= <mat646 at gmail.com>
Message-ID:
In-Reply-To: <llvm.org/llvm/llvm-project/pull/127825 at github.com>


================
@@ -2733,7 +2734,8 @@ void CombinerHelper::applyCombineTruncOfShift(
 bool CombinerHelper::matchAnyExplicitUseIsUndef(MachineInstr &MI) const {
   return any_of(MI.explicit_uses(), [this](const MachineOperand &MO) {
     return MO.isReg() &&
-           getOpcodeDef(TargetOpcode::G_IMPLICIT_DEF, MO.getReg(), MRI);
+           (getOpcodeDef(TargetOpcode::G_IMPLICIT_DEF, MO.getReg(), MRI) ||
+            getOpcodeDef(TargetOpcode::G_POISON, MO.getReg(), MRI));
----------------
arsenm wrote:

Can you add a fixme? We shouldn't look up copy chains twice just to check 2 opcodes. We should delete getOpcodeDef as a helper, it's not good 

https://github.com/llvm/llvm-project/pull/127825


More information about the llvm-commits mailing list