From: Tim Northover <tnorthover@apple.com>
Date: Tue, 27 Aug 2019 10:21:11 +0000 (+0000)
Subject: AArch64: avoid creating cycle in DAG for post-increment NEON ops.
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=0b15c50b407d3d21b0cda46d50fe6da303258cd3;p=llvm

AArch64: avoid creating cycle in DAG for post-increment NEON ops.

Inserting a value into Visited has the effect of terminating a search for
predecessors if that node is seen. This is legitimate for the base address, and
acts as a slight performance optimization, but the vector-building node can be
paert of a legitimate cycle so we shouldn't stop searching there.

PR43056.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@370036 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/AArch64/AArch64ISelLowering.cpp b/lib/Target/AArch64/AArch64ISelLowering.cpp
index be248ee898c..6d2f363858e 100644
--- a/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -10694,7 +10694,7 @@ static SDValue performPostLD1Combine(SDNode *N,
     // are predecessors to each other or the Vector.
     SmallPtrSet<const SDNode *, 32> Visited;
     SmallVector<const SDNode *, 16> Worklist;
-    Visited.insert(N);
+    Visited.insert(Addr.getNode());
     Worklist.push_back(User);
     Worklist.push_back(LD);
     Worklist.push_back(Vector.getNode());
diff --git a/test/CodeGen/AArch64/arm64-indexed-vector-ldst.ll b/test/CodeGen/AArch64/arm64-indexed-vector-ldst.ll
index 8b6a4cae7ed..f3ac9b21f53 100644
--- a/test/CodeGen/AArch64/arm64-indexed-vector-ldst.ll
+++ b/test/CodeGen/AArch64/arm64-indexed-vector-ldst.ll
@@ -6319,3 +6319,22 @@ define void  @test_ld1lane_build_i8(i8* %a, i8* %b, i8* %c, i8* %d, i8* %e, i8*
   store <8 x i8> %sub, <8 x i8>* %p
   ret void
 }
+
+define <4 x i32> @test_inc_cycle(<4 x i32> %vec, i32* %in) {
+; CHECK-LABEL: test_inc_cycle:
+; CHECK: ld1.s { v0 }[0], [x0]{{$}}
+
+  %elt = load i32, i32* %in
+  %newvec = insertelement <4 x i32> %vec, i32 %elt, i32 0
+
+  ; %inc cannot be %elt directly because we check that the load is only
+  ; used by the insert before trying to form post-inc.
+  %inc.vec = bitcast <4 x i32> %newvec to <2 x i64>
+  %inc = extractelement <2 x i64> %inc.vec, i32 0
+  %newaddr = getelementptr i32, i32* %in, i64 %inc
+  store i32* %newaddr, i32** @var
+
+  ret <4 x i32> %newvec
+}
+
+@var = global i32* null