From aac455af997edb35b244a8a19af99a64d9ddc93e Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Sat, 14 Jun 2014 04:26:05 +0000
Subject: [PATCH] R600: Fix asserts related to constant initializers

This would assert if a constant address space was extern
and therefore didn't have an initializer. If the initializer
was undef, it would hit the unreachable unhandled initializer case.

An extern global should never really occur since we don't have
machine linking, but bugpoint likes to remove initializers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210967 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/R600/AMDGPUISelLowering.cpp        | 25 +++++++++++++++----
 .../R600/no-initializer-constant-addrspace.ll | 20 +++++++++++++++
 2 files changed, 40 insertions(+), 5 deletions(-)
 create mode 100644 test/CodeGen/R600/no-initializer-constant-addrspace.ll

diff --git a/lib/Target/R600/AMDGPUISelLowering.cpp b/lib/Target/R600/AMDGPUISelLowering.cpp
index 3ea789da5aa..7867de23fd2 100644
--- a/lib/Target/R600/AMDGPUISelLowering.cpp
+++ b/lib/Target/R600/AMDGPUISelLowering.cpp
@@ -655,6 +655,14 @@ SDValue AMDGPUTargetLowering::LowerConstantInitializer(const Constant* Init,
     return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Chains);
   }
 
+  if (isa<UndefValue>(Init)) {
+    EVT VT = EVT::getEVT(InitTy);
+    PointerType *PtrTy = PointerType::get(InitTy, AMDGPUAS::PRIVATE_ADDRESS);
+    return DAG.getStore(Chain, DL, DAG.getUNDEF(VT), InitPtr,
+                        MachinePointerInfo(UndefValue::get(PtrTy)), false, false,
+                        TD->getPrefTypeAlignment(InitTy));
+  }
+
   Init->dump();
   llvm_unreachable("Unhandled constant initializer");
 }
@@ -694,11 +702,19 @@ SDValue AMDGPUTargetLowering::LowerGlobalAddress(AMDGPUMachineFunction* MFI,
     unsigned Size = TD->getTypeAllocSize(EltType);
     unsigned Alignment = TD->getPrefTypeAlignment(EltType);
 
+    MVT PrivPtrVT = getPointerTy(AMDGPUAS::PRIVATE_ADDRESS);
+    MVT ConstPtrVT = getPointerTy(AMDGPUAS::CONSTANT_ADDRESS);
+
+    int FI = FrameInfo->CreateStackObject(Size, Alignment, false);
+    SDValue InitPtr = DAG.getFrameIndex(FI, PrivPtrVT);
+
     const GlobalVariable *Var = cast<GlobalVariable>(GV);
+    if (!Var->hasInitializer()) {
+      // This has no use, but bugpoint will hit it.
+      return DAG.getZExtOrTrunc(InitPtr, SDLoc(Op), ConstPtrVT);
+    }
+
     const Constant *Init = Var->getInitializer();
-    int FI = FrameInfo->CreateStackObject(Size, Alignment, false);
-    SDValue InitPtr = DAG.getFrameIndex(FI,
-        getPointerTy(AMDGPUAS::PRIVATE_ADDRESS));
     SmallVector<SDNode*, 8> WorkList;
 
     for (SDNode::use_iterator I = DAG.getEntryNode()->use_begin(),
@@ -717,8 +733,7 @@ SDValue AMDGPUTargetLowering::LowerGlobalAddress(AMDGPUMachineFunction* MFI,
       }
       DAG.UpdateNodeOperands(*I, Ops);
     }
-    return DAG.getZExtOrTrunc(InitPtr, SDLoc(Op),
-        getPointerTy(AMDGPUAS::CONSTANT_ADDRESS));
+    return DAG.getZExtOrTrunc(InitPtr, SDLoc(Op), ConstPtrVT);
   }
   }
 }
diff --git a/test/CodeGen/R600/no-initializer-constant-addrspace.ll b/test/CodeGen/R600/no-initializer-constant-addrspace.ll
new file mode 100644
index 00000000000..ab82e7ee799
--- /dev/null
+++ b/test/CodeGen/R600/no-initializer-constant-addrspace.ll
@@ -0,0 +1,20 @@
+; RUN: llc -march=r600 -mcpu=SI -o /dev/null %s
+; RUN: llc -march=r600 -mcpu=cypress -o /dev/null %s
+
+@extern_const_addrspace = external unnamed_addr addrspace(2) constant [5 x i32], align 4
+
+; FUNC-LABEL: @load_extern_const_init
+define void @load_extern_const_init(i32 addrspace(1)* %out) nounwind {
+  %val = load i32 addrspace(2)* getelementptr ([5 x i32] addrspace(2)* @extern_const_addrspace, i64 0, i64 3), align 4
+  store i32 %val, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
+@undef_const_addrspace = unnamed_addr addrspace(2) constant [5 x i32] undef, align 4
+
+; FUNC-LABEL: @load_undef_const_init
+define void @load_undef_const_init(i32 addrspace(1)* %out) nounwind {
+  %val = load i32 addrspace(2)* getelementptr ([5 x i32] addrspace(2)* @undef_const_addrspace, i64 0, i64 3), align 4
+  store i32 %val, i32 addrspace(1)* %out, align 4
+  ret void
+}
-- 
2.34.1