[llvm] [PowerPC] Use zext instead of anyext in custom and combine (PR #68784)
Nikita Popov via llvm-commits
llvm-commits at lists.llvm.org
Wed Oct 11 03:21:03 PDT 2023
https://github.com/nikic created https://github.com/llvm/llvm-project/pull/68784
This custom combine currently converts `and(anyext(x),c)` into `anyext(and(x,c))`. This is not correct, because the original expression guaranteed that the high bits are zero, while the new one sets them to undef.
Emit `zext(and(x,c))` instead.
Fixes https://github.com/llvm/llvm-project/issues/68783.
>From c3b6edd1d816b7f0ba5aa2dc1ab44088027f69e5 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov at redhat.com>
Date: Wed, 11 Oct 2023 12:16:04 +0200
Subject: [PATCH] [PowerPC] Use zext instead of anyext in custom and combine
This custom combine currently converts an `and(anyext(x),c)`
into `anyext(and(x,c))`. This is not correct, because the original
expression guaranteed that the high bits are zero, while the new
one sets them to undef.
Emit `zext(and(x,c))` instead.
Fixes https://github.com/llvm/llvm-project/issues/68783.
---
llvm/lib/Target/PowerPC/PPCISelLowering.cpp | 2 +-
llvm/test/CodeGen/PowerPC/and-extend-combine.ll | 3 ++-
2 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index c6257dcdf76f633..5e0c2d62f5a9cb5 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -15588,7 +15588,7 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N,
break;
SDValue ConstOp = DAG.getConstant(Imm, dl, MVT::i32);
SDValue NarrowAnd = DAG.getNode(ISD::AND, dl, MVT::i32, NarrowOp, ConstOp);
- return DAG.getAnyExtOrTrunc(NarrowAnd, dl, N->getValueType(0));
+ return DAG.getZExtOrTrunc(NarrowAnd, dl, N->getValueType(0));
}
case ISD::SHL:
return combineSHL(N, DCI);
diff --git a/llvm/test/CodeGen/PowerPC/and-extend-combine.ll b/llvm/test/CodeGen/PowerPC/and-extend-combine.ll
index 11f83ccc6806d0d..0b30d48912dc374 100644
--- a/llvm/test/CodeGen/PowerPC/and-extend-combine.ll
+++ b/llvm/test/CodeGen/PowerPC/and-extend-combine.ll
@@ -23,11 +23,12 @@ bb:
ret ptr %i8
}
-; FIXME: This is a miscompile.
define void @pr68783(i32 %x, ptr %p) {
; CHECK-LABEL: pr68783:
; CHECK: # %bb.0:
; CHECK-NEXT: rlwinm r3, r3, 31, 24, 31
+; CHECK-NEXT: li r5, 0
+; CHECK-NEXT: sth r5, 4(r4)
; CHECK-NEXT: stw r3, 0(r4)
; CHECK-NEXT: blr
%lshr = lshr i32 %x, 1
More information about the llvm-commits
mailing list