From 43d225dc8bdaa37e1b0a222f7510ade18a7267e4 Mon Sep 17 00:00:00 2001
From: Kalle Raiskila <kalle.raiskila@nokia.com>
Date: Wed, 9 Jun 2010 09:58:17 +0000
Subject: [PATCH] Fix SPU to cope with vector insertelement to an undef
 position. We default to inserting to lane 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105722 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/CellSPU/SPUISelLowering.cpp | 11 ++++++++---
 test/CodeGen/CellSPU/vecinsert.ll      | 15 +++++++++++++--
 2 files changed, 21 insertions(+), 5 deletions(-)

diff --git a/lib/Target/CellSPU/SPUISelLowering.cpp b/lib/Target/CellSPU/SPUISelLowering.cpp
index 081e8d0db0e..affd066876f 100644
--- a/lib/Target/CellSPU/SPUISelLowering.cpp
+++ b/lib/Target/CellSPU/SPUISelLowering.cpp
@@ -2056,14 +2056,19 @@ static SDValue LowerINSERT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) {
   DebugLoc dl = Op.getDebugLoc();
   EVT VT = Op.getValueType();
 
-  ConstantSDNode *CN = cast<ConstantSDNode>(IdxOp);
-  assert(CN != 0 && "LowerINSERT_VECTOR_ELT: Index is not constant!");
+  // use 0 when the lane to insert to is 'undef'
+  int64_t Idx=0;
+  if (IdxOp.getOpcode() != ISD::UNDEF) {
+    ConstantSDNode *CN = cast<ConstantSDNode>(IdxOp);
+    assert(CN != 0 && "LowerINSERT_VECTOR_ELT: Index is not constant!");
+    Idx = (CN->getSExtValue());
+  }
 
   EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy();
   // Use $sp ($1) because it's always 16-byte aligned and it's available:
   SDValue Pointer = DAG.getNode(SPUISD::IndirectAddr, dl, PtrVT,
                                 DAG.getRegister(SPU::R1, PtrVT),
-                                DAG.getConstant(CN->getSExtValue(), PtrVT));
+                                DAG.getConstant(Idx, PtrVT));
   SDValue ShufMask = DAG.getNode(SPUISD::SHUFFLE_MASK, dl, VT, Pointer);
 
   SDValue result =
diff --git a/test/CodeGen/CellSPU/vecinsert.ll b/test/CodeGen/CellSPU/vecinsert.ll
index 9a00c1f29f8..8dcab1d84c9 100644
--- a/test/CodeGen/CellSPU/vecinsert.ll
+++ b/test/CodeGen/CellSPU/vecinsert.ll
@@ -1,17 +1,19 @@
 ; RUN: llc < %s -march=cellspu > %t1.s
 ; RUN: grep cbd     %t1.s | count 5
 ; RUN: grep chd     %t1.s | count 5
-; RUN: grep cwd     %t1.s | count 10
+; RUN: grep cwd     %t1.s | count 11
 ; RUN: grep -w il   %t1.s | count 5
 ; RUN: grep -w ilh  %t1.s | count 6
 ; RUN: grep iohl    %t1.s | count 1
 ; RUN: grep ilhu    %t1.s | count 4
-; RUN: grep shufb   %t1.s | count 26
+; RUN: grep shufb   %t1.s | count 27
 ; RUN: grep 17219   %t1.s | count 1 
 ; RUN: grep 22598   %t1.s | count 1
 ; RUN: grep -- -39  %t1.s | count 1
 ; RUN: grep    24   %t1.s | count 1
 ; RUN: grep  1159   %t1.s | count 1
+; RUN: FileCheck %s < %t1.s
+
 ; ModuleID = 'vecinsert.bc'
 target datalayout = "E-p:32:32:128-f64:64:128-f32:32:128-i64:32:128-i32:32:128-i16:16:128-i8:8:128-i1:8:128-a0:0:128-v128:128:128"
 target triple = "spu-unknown-elf"
@@ -118,3 +120,12 @@ entry:
 	store <2 x double> %tmp3, <2 x double>* %arrayidx
 	ret void
 }
+
+define <4 x i32> @undef_v4i32( i32 %param ) {
+	;CHECK: cwd
+	;CHECK: lqa
+	;CHECK: shufb
+	%val = insertelement <4 x i32> <i32 1, i32 2, i32 3, i32 4>, i32 %param, i32 undef 
+	ret <4 x i32> %val
+}
+