AMDGPU: Treat 0 as private null pointer in addrspacecast lowering

llvm-svn: 297658

4 files changed, 23 insertions, 17 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.h b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.h
index bf0b954..e36e940 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.h
@@ -62,7 +62,6 @@ public:
   /// Get the integer value of a null pointer in the given address space.
   uint64_t getNullPointerValue(unsigned AddrSpace) const {
     switch(AddrSpace) {
-    case AMDGPUAS::PRIVATE_ADDRESS:
     case AMDGPUAS::LOCAL_ADDRESS:
     case AMDGPUAS::REGION_ADDRESS:
       return -1;
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index f6a4dc5..8a621952 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2374,21 +2374,25 @@ SDValue SITargetLowering::getSegmentAperture(unsigned AS,
                          MachineMemOperand::MOInvariant);
 }
 
+// FIXME: Really support non-0 null pointers.
+static int getSegmentNullPtrValue(unsigned AS) {
+  return AS == AMDGPUAS::LOCAL_ADDRESS ? -1 : 0;
+}
+
 SDValue SITargetLowering::lowerADDRSPACECAST(SDValue Op,
                                              SelectionDAG &DAG) const {
   SDLoc SL(Op);
   const AddrSpaceCastSDNode *ASC = cast<AddrSpaceCastSDNode>(Op);
 
   SDValue Src = ASC->getOperand(0);
-
-  // FIXME: Really support non-0 null pointers.
-  SDValue SegmentNullPtr = DAG.getConstant(-1, SL, MVT::i32);
   SDValue FlatNullPtr = DAG.getConstant(0, SL, MVT::i64);
 
   // flat -> local/private
   if (ASC->getSrcAddressSpace() == AMDGPUAS::FLAT_ADDRESS) {
-    if (ASC->getDestAddressSpace() == AMDGPUAS::LOCAL_ADDRESS ||
-        ASC->getDestAddressSpace() == AMDGPUAS::PRIVATE_ADDRESS) {
+    unsigned DestAS = ASC->getDestAddressSpace();
+    if (DestAS == AMDGPUAS::LOCAL_ADDRESS || DestAS == AMDGPUAS::PRIVATE_ADDRESS) {
+      SDValue SegmentNullPtr
+        = DAG.getConstant(getSegmentNullPtrValue(DestAS), SL, MVT::i32);
       SDValue NonNull = DAG.getSetCC(SL, MVT::i1, Src, FlatNullPtr, ISD::SETNE);
       SDValue Ptr = DAG.getNode(ISD::TRUNCATE, SL, MVT::i32, Src);
 
@@ -2399,8 +2403,11 @@ SDValue SITargetLowering::lowerADDRSPACECAST(SDValue Op,
 
   // local/private -> flat
   if (ASC->getDestAddressSpace() == AMDGPUAS::FLAT_ADDRESS) {
-    if (ASC->getSrcAddressSpace() == AMDGPUAS::LOCAL_ADDRESS ||
-        ASC->getSrcAddressSpace() == AMDGPUAS::PRIVATE_ADDRESS) {
+    unsigned SrcAS = ASC->getSrcAddressSpace();
+    if (SrcAS == AMDGPUAS::LOCAL_ADDRESS || SrcAS == AMDGPUAS::PRIVATE_ADDRESS) {
+      SDValue SegmentNullPtr
+        = DAG.getConstant(getSegmentNullPtrValue(SrcAS), SL, MVT::i32);
+
       SDValue NonNull
         = DAG.getSetCC(SL, MVT::i1, Src, SegmentNullPtr, ISD::SETNE);
 
diff --git a/llvm/test/CodeGen/AMDGPU/addrspacecast.ll b/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
index 12dcda9..b6ada5e 100644
--- a/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
+++ b/llvm/test/CodeGen/AMDGPU/addrspacecast.ll
@@ -49,7 +49,7 @@ define void @use_group_to_flat_addrspacecast(i32 addrspace(3)* %ptr) #0 {
 
 ; HSA-DAG: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[PTR]]
 
-; HSA-DAG: v_cmp_ne_u32_e64 vcc, [[PTR]], -1
+; HSA-DAG: v_cmp_ne_u32_e64 vcc, [[PTR]], 0
 ; HSA-DAG: v_cndmask_b32_e32 v[[HI:[0-9]+]], 0, [[VAPERTURE]]
 ; HSA-DAG: v_cndmask_b32_e32 v[[LO:[0-9]+]], 0, [[VPTR]]
 ; HSA-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7
@@ -116,7 +116,7 @@ define void @use_flat_to_group_addrspacecast(i32 addrspace(4)* %ptr) #0 {
 ; HSA: s_load_dwordx2 s{{\[}}[[PTR_LO:[0-9]+]]:[[PTR_HI:[0-9]+]]{{\]}}
 ; HSA-DAG: v_cmp_ne_u64_e64 vcc, s{{\[}}[[PTR_LO]]:[[PTR_HI]]{{\]}}, 0{{$}}
 ; HSA-DAG: v_mov_b32_e32 v[[VPTR_LO:[0-9]+]], s[[PTR_LO]]
-; HSA-DAG: v_cndmask_b32_e32 [[CASTPTR:v[0-9]+]], -1, v[[VPTR_LO]]
+; HSA-DAG: v_cndmask_b32_e32 [[CASTPTR:v[0-9]+]], 0, v[[VPTR_LO]]
 ; HSA-DAG: v_mov_b32_e32 v[[K:[0-9]+]], 0{{$}}
 ; HSA: buffer_store_dword v[[K]], [[CASTPTR]], s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}}
 define void @use_flat_to_private_addrspacecast(i32 addrspace(4)* %ptr) #0 {
@@ -195,14 +195,14 @@ define void @cast_neg1_flat_to_group_addrspacecast() #0 {
   ret void
 }
 
+; FIXME: Shouldn't need to enable queue ptr
 ; HSA-LABEL: {{^}}cast_0_private_to_flat_addrspacecast:
-; CI: s_load_dword [[APERTURE:s[0-9]+]], s[4:5], 0x11
-; CI-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], [[APERTURE]]
-
-; GFX9-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], src_private_base
+; CI: enable_sgpr_queue_ptr = 1
+; GFX9: enable_sgpr_queue_ptr = 0
 
 ; HSA-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], 0{{$}}
 ; HSA-DAG: v_mov_b32_e32 v[[K:[0-9]+]], 7{{$}}
+; HSA: v_mov_b32_e32 v[[HI:[0-9]+]], 0{{$}}
 ; HSA: flat_store_dword v{{\[}}[[LO]]:[[HI]]{{\]}}, v[[K]]
 define void @cast_0_private_to_flat_addrspacecast() #0 {
   %cast = addrspacecast i32* null to i32 addrspace(4)*
@@ -211,7 +211,7 @@ define void @cast_0_private_to_flat_addrspacecast() #0 {
 }
 
 ; HSA-LABEL: {{^}}cast_0_flat_to_private_addrspacecast:
-; HSA-DAG: v_mov_b32_e32 [[PTR:v[0-9]+]], -1{{$}}
+; HSA-DAG: v_mov_b32_e32 [[PTR:v[0-9]+]], 0{{$}}
 ; HSA-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 7{{$}}
 ; HSA: buffer_store_dword [[K]], [[PTR]], s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen
 define void @cast_0_flat_to_private_addrspacecast() #0 {
diff --git a/llvm/test/CodeGen/AMDGPU/nullptr.ll b/llvm/test/CodeGen/AMDGPU/nullptr.ll
index b4bdf71..0df16da 100644
--- a/llvm/test/CodeGen/AMDGPU/nullptr.ll
+++ b/llvm/test/CodeGen/AMDGPU/nullptr.ll
@@ -3,7 +3,7 @@
 %struct.S = type { i32*, i32 addrspace(1)*, i32 addrspace(2)*, i32 addrspace(3)*, i32 addrspace(4)*, i32 addrspace(5)*}
 
 ; CHECK-LABEL: nullptr_priv:
-; CHECK-NEXT: .long -1
+; CHECK-NEXT: .long 0
 @nullptr_priv = global i32* addrspacecast (i32 addrspace(4)* null to i32*)
 
 ; CHECK-LABEL: nullptr_glob:
@@ -95,7 +95,7 @@
 @nullptr23 = global i32 addrspace(23)* addrspacecast (i32 addrspace(4)* null to i32 addrspace(23)*)
 
 ; CHECK-LABEL: structWithPointers:
-; CHECK-NEXT: .long -1
+; CHECK-NEXT: .long 0
 ; CHECK-NEXT: .zero 4
 ; CHECK-NEXT: .quad 0
 ; CHECK-NEXT: .quad 0