[WebAssembly] Tidy up handling of global symbol relocations

[ncw]

Summary

Currently global relocations are using their own fixup kind, which is odd since it's just a uleb32. That's because we're not tracking the difference between data and global symbols, which would enable us to process the fixup in the same way as all the other fixups.

Also, we're using some target flags already (like 0x1 to mean "function") which are just defined inline in the code and not in a header!

---

These changes are preliminary to the next commit which will add TLS (__thread_pointer) support.

I'm new to this area of the codebase, this might be all wrong! I think it looks sensible so far though.

[ncw]

lib/MC/MCWasmStreamer.cpp

120

This is the only bit I was really unsure about, in terms of whether it breaks anything... currently the "ELF_TypeObject" attribute is set for both WASM_SYMBOL_TYPE_DATA and WASM_SYMBOL_TYPE_GLOBAL symbols, which is a bit odd. It seems to come from higher-level code that just sets it by default, rather than because we really want that attribute there in our Wasm objects.

[sbc100]

Looks good to me. @sunfish knows this stuff better than me. I guess the .s file form for -elf won't change at all (since we don't use globals there).. otherwise we could need to fix binaryen (which currently consumes .s).

lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTargetDesc.h

91

Other platforms seem to call this TOF. Any reason not to follow suit?

[ncw]

Thanks for the review Sam, I hadn't added you to this one you wouldn't be bothered by all my commits!

Things are really pretty snowy here, which is unusual and nice, although quite inconvenient. Have a look at some pics of Britain in the news if you haven't, it's lovely and white.

lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTargetDesc.h

91

If it's inconsistent, it's sheer ignorance! I think I did copy this name from one of the other targets, maybe the wrong one though.

I'll check and fix the name.

[sbc100]

In D44030#1025941, @ncw wrote:

Thanks for the review Sam, I hadn't added you to this one you wouldn't be bothered by all my commits!

Things are really pretty snowy here, which is unusual and nice, although quite inconvenient. Have a look at some pics of Britain in the news if you haven't, it's lovely and white.

Oh.. I must have been auto-cc'd by herald. I didn't notice it was just a CC.

I heard its cold there yes. We are also getting a lot of snow up in Tahoe which is great news for California.

[ncw]

Ping - has anyone who's worked on this code got a chance to look at it this week? Thanks very much!

[ncw]

Ping again...

[ncw]

@sunfish, @dschuff, it's been another fortnight on this PR. If you don't want to look at it, or want to suggest another reviewer - that's fine, but it would be helpful to know, to get some feedback on whether I should by trying to get this in, or abandon it.

The main improvement in this PR is removing the strange WebAssembly::fixup_code_global_index fixup kind, and switching to using plain uleb32 fixups for global indexes, just the same as function index fixups currently do. Also I've pulled out some hardcoded constants into headers, and added an operand flag for globals to match the existing operand flag for function indexes (this is necessary to enable using the generic uleb32 fixup kind with globals).

[sunfish]

First, I apologize, I'm not sure how I lost track of this patch and didn't see the pings.

The patch looks good to me. Adding @GLOBAL rather than inferring it from the fact that a symbol appears in a get_global/set_global seems reasonable.

include/llvm/MC/MCExpr.h

282

Please vertically align the comment here.

[ncw]

Updated with Dan and Sam's review comments.

Thanks very much for looking at this!

include/llvm/MC/MCExpr.h

enum VariantKind : uint16_t {
     VK_Hexagon_GD_GOT,
     VK_Hexagon_LD_GOT,
     VK_Hexagon_GD_PLT,
     VK_Hexagon_LD_PLT,
     VK_Hexagon_IE,
     VK_Hexagon_IE_GOT,
 
     VK_WebAssembly_FUNCTION, // Function table index, rather than virtual addr
+    VK_WebAssembly_GLOBAL, // Global object index
     VK_WebAssembly_TYPEINDEX,// Type table index
 
     VK_AMDGPU_GOTPCREL32_LO, // symbol@gotpcrel32@lo
     VK_AMDGPU_GOTPCREL32_HI, // symbol@gotpcrel32@hi
     VK_AMDGPU_REL32_LO,      // symbol@rel32@lo
     VK_AMDGPU_REL32_HI,      // symbol@rel32@hi
 
     VK_TPREL,

lib/MC/MCExpr.cpp

StringRef MCSymbolRefExpr::getVariantKindName(VariantKind Kind) {
   case VK_Hexagon_GPREL: return "GPREL";
   case VK_Hexagon_GD_GOT: return "GDGOT";
   case VK_Hexagon_LD_GOT: return "LDGOT";
   case VK_Hexagon_GD_PLT: return "GDPLT";
   case VK_Hexagon_LD_PLT: return "LDPLT";
   case VK_Hexagon_IE: return "IE";
   case VK_Hexagon_IE_GOT: return "IEGOT";
   case VK_WebAssembly_FUNCTION: return "FUNCTION";
+  case VK_WebAssembly_GLOBAL: return "GLOBAL";
   case VK_WebAssembly_TYPEINDEX: return "TYPEINDEX";
   case VK_AMDGPU_GOTPCREL32_LO: return "gotpcrel32@lo";
   case VK_AMDGPU_GOTPCREL32_HI: return "gotpcrel32@hi";
   case VK_AMDGPU_REL32_LO: return "rel32@lo";
   case VK_AMDGPU_REL32_HI: return "rel32@hi";
   }
   llvm_unreachable("Invalid variant kind");
 }

lib/MC/MCWasmStreamer.cpp

case MCSA_Global:
     Symbol->setExternal(true);
     break;
 
   case MCSA_ELF_TypeFunction:
     Symbol->setType(wasm::WASM_SYMBOL_TYPE_FUNCTION);
     break;
 
   case MCSA_ELF_TypeObject:
-    Symbol->setType(wasm::WASM_SYMBOL_TYPE_DATA);
     break;
 
   default:
     // unrecognized directive
     llvm_unreachable("unexpected MCSymbolAttr");
     return false;
   }
 

lib/Target/WebAssembly/MCTargetDesc/WebAssemblyFixupKinds.h

 
 namespace llvm {
 namespace WebAssembly {
 enum Fixups {
   fixup_code_sleb128_i32 = FirstTargetFixupKind,      // 32-bit signed
   fixup_code_sleb128_i64,                             // 64-bit signed
   fixup_code_uleb128_i32,                             // 32-bit unsigned
 
-  fixup_code_global_index,                            // 32-bit unsigned
-
   // Marker
   LastTargetFixupKind,
   NumTargetFixupKinds = LastTargetFixupKind - FirstTargetFixupKind
 };
 } // end namespace WebAssembly
 } // end namespace llvm
 
 #endif

lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCCodeEmitter.cpp

if (MO.isReg()) {
       } else if (Info.OperandType == WebAssembly::OPERAND_I64IMM) {
         FixupKind = MCFixupKind(WebAssembly::fixup_code_sleb128_i64);
         PaddedSize = 10;
       } else if (Info.OperandType == WebAssembly::OPERAND_FUNCTION32 ||
                  Info.OperandType == WebAssembly::OPERAND_OFFSET32 ||
                  Info.OperandType == WebAssembly::OPERAND_TYPEINDEX) {
         FixupKind = MCFixupKind(WebAssembly::fixup_code_uleb128_i32);
       } else if (Info.OperandType == WebAssembly::OPERAND_GLOBAL) {
-        FixupKind = MCFixupKind(WebAssembly::fixup_code_global_index);
+        FixupKind = MCFixupKind(WebAssembly::fixup_code_uleb128_i32);
       } else {
         llvm_unreachable("unexpected symbolic operand kind");
       }
       Fixups.push_back(MCFixup::create(
           OS.tell() - Start, MO.getExpr(),
           FixupKind, MI.getLoc()));
       ++MCNumFixups;
       encodeULEB128(0, OS, PaddedSize);

lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTargetDesc.h

 enum {
   // For variadic instructions, this flag indicates whether an operand
   // in the variable_ops range is an immediate value.
   VariableOpIsImmediate = (1 << 0),
   // For immediate values in the variable_ops range, this flag indicates
   // whether the value represents a control-flow label.
   VariableOpImmediateIsLabel = (1 << 1)
 };
+
+// WebAssembly-specific MachineOperand target flags.
+enum WebAssemblyMOTargetFlagVal {
+  MO_NO_FLAG = 0,
+
+  // Flags to indicate the type of the symbol being referenced
+  MO_SYMBOL_FUNCTION = 0x1,
+  MO_SYMBOL_GLOBAL   = 0x2,
+  MO_SYMBOL_MASK     = 0x3,
+};
 } // end namespace WebAssemblyII
 
 } // end namespace llvm
 
 // Defines symbolic names for WebAssembly registers. This defines a mapping from
 // register name to register number.
 //
 #define GET_REGINFO_ENUM

lib/Target/WebAssembly/MCTargetDesc/WebAssemblyWasmObjectWriter.cpp

static bool IsFunctionExpr(const MCExpr *Expr) {
   return false;
 }
 
 static bool IsFunctionType(const MCValue &Target) {
   const MCSymbolRefExpr *RefA = Target.getSymA();
   return RefA && RefA->getKind() == MCSymbolRefExpr::VK_WebAssembly_TYPEINDEX;
 }
 
+static bool IsGlobalType(const MCValue &Target) {
+  const MCSymbolRefExpr *RefA = Target.getSymA();
+  return RefA && RefA->getKind() == MCSymbolRefExpr::VK_WebAssembly_GLOBAL;
+}
+
 unsigned
 WebAssemblyWasmObjectWriter::getRelocType(const MCValue &Target,
                                           const MCFixup &Fixup) const {
   // WebAssembly functions are not allocated in the data address space. To
   // resolve a pointer to a function, we must use a special relocation type.
   bool IsFunction = IsFunctionExpr(Fixup.getValue());
 
   switch (unsigned(Fixup.getKind())) {
-  case WebAssembly::fixup_code_global_index:
-    return wasm::R_WEBASSEMBLY_GLOBAL_INDEX_LEB;
   case WebAssembly::fixup_code_sleb128_i32:
     if (IsFunction)
       return wasm::R_WEBASSEMBLY_TABLE_INDEX_SLEB;
     return wasm::R_WEBASSEMBLY_MEMORY_ADDR_SLEB;
   case WebAssembly::fixup_code_sleb128_i64:
     llvm_unreachable("fixup_sleb128_i64 not implemented yet");
   case WebAssembly::fixup_code_uleb128_i32:
+    if (IsGlobalType(Target))
+      return wasm::R_WEBASSEMBLY_GLOBAL_INDEX_LEB;
     if (IsFunctionType(Target))
       return wasm::R_WEBASSEMBLY_TYPE_INDEX_LEB;
     if (IsFunction)
       return wasm::R_WEBASSEMBLY_FUNCTION_INDEX_LEB;
     return wasm::R_WEBASSEMBLY_MEMORY_ADDR_LEB;
   case FK_Data_4:
     if (IsFunction)
       return wasm::R_WEBASSEMBLY_TABLE_INDEX_I32;

lib/Target/WebAssembly/WebAssemblyFrameLowering.cpp

if (MF.getSubtarget<WebAssemblySubtarget>()
     BuildMI(MBB, InsertStore, DL, TII->get(WebAssembly::STORE_I32))
         .addImm(2)  // p2align
         .addExternalSymbol(SPSymbol)
         .addReg(Zero)
         .addReg(SrcReg)
         .addMemOperand(MMO);
   } else {
     BuildMI(MBB, InsertStore, DL, TII->get(WebAssembly::SET_GLOBAL_I32))
-        .addExternalSymbol(SPSymbol)
+        .addExternalSymbol(SPSymbol, WebAssemblyII::MO_SYMBOL_GLOBAL)
         .addReg(SrcReg);
   }
 }
 
 MachineBasicBlock::iterator
 WebAssemblyFrameLowering::eliminateCallFramePseudoInstr(
     MachineFunction &MF, MachineBasicBlock &MBB,
     MachineBasicBlock::iterator I) const {
if (MF.getSubtarget<WebAssemblySubtarget>()
     // Load the SP value.
     BuildMI(MBB, InsertPt, DL, TII->get(WebAssembly::LOAD_I32), SPReg)
         .addImm(2)       // p2align
         .addExternalSymbol(SPSymbol)
         .addReg(Zero)    // addr
         .addMemOperand(LoadMMO);
   } else {
     BuildMI(MBB, InsertPt, DL, TII->get(WebAssembly::GET_GLOBAL_I32), SPReg)
-        .addExternalSymbol(SPSymbol);
+        .addExternalSymbol(SPSymbol, WebAssemblyII::MO_SYMBOL_GLOBAL);
   }
 
   bool HasBP = hasBP(MF);
   if (HasBP) {
     auto FI = MF.getInfo<WebAssemblyFunctionInfo>();
     unsigned BasePtr = MRI.createVirtualRegister(PtrRC);
     FI->setBasePointerVreg(BasePtr);
     BuildMI(MBB, InsertPt, DL, TII->get(WebAssembly::COPY), BasePtr)

lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

assert(ES->getTargetFlags() == 0 &&
          "Unexpected target flags on generic ExternalSymbolSDNode");
   // Set the TargetFlags to 0x1 which indicates that this is a "function"
   // symbol rather than a data symbol. We do this unconditionally even though
   // we don't know anything about the symbol other than its name, because all
   // external symbols used in target-independent SelectionDAG code are for
   // functions.
   return DAG.getNode(WebAssemblyISD::Wrapper, DL, VT,
                      DAG.getTargetExternalSymbol(ES->getSymbol(), VT,
-                                                 /*TargetFlags=*/0x1));
+                                                 WebAssemblyII::MO_SYMBOL_FUNCTION));
 }
 
 SDValue WebAssemblyTargetLowering::LowerJumpTable(SDValue Op,
                                                   SelectionDAG &DAG) const {
   // There's no need for a Wrapper node because we always incorporate a jump
   // table operand into a BR_TABLE instruction, rather than ever
   // materializing it in a register.
   const JumpTableSDNode *JT = cast<JumpTableSDNode>(Op);

lib/Target/WebAssembly/WebAssemblyMCInstLower.h

 /// This class is used to lower an MachineInstr into an MCInst.
 class LLVM_LIBRARY_VISIBILITY WebAssemblyMCInstLower {
   MCContext &Ctx;
   WebAssemblyAsmPrinter &Printer;
 
   MCSymbol *GetGlobalAddressSymbol(const MachineOperand &MO) const;
   MCSymbol *GetExternalSymbolSymbol(const MachineOperand &MO) const;
   MCOperand LowerSymbolOperand(MCSymbol *Sym, int64_t Offset,
-                               bool IsFunc) const;
+                               bool IsFunc, bool IsGlob) const;
 
 public:
   WebAssemblyMCInstLower(MCContext &ctx, WebAssemblyAsmPrinter &printer)
       : Ctx(ctx), Printer(printer) {}
   void Lower(const MachineInstr *MI, MCInst &OutMI) const;
 };
 } // end namespace llvm
 
 #endif

lib/Target/WebAssembly/WebAssemblyMCInstLower.cpp

MCSymbol *WebAssemblyMCInstLower::GetExternalSymbolSymbol(
   WasmSym->setParams(std::move(Params));
   WasmSym->setType(wasm::WASM_SYMBOL_TYPE_FUNCTION);
 
   return WasmSym;
 }
 
 MCOperand WebAssemblyMCInstLower::LowerSymbolOperand(MCSymbol *Sym,
                                                      int64_t Offset,
-                                                     bool IsFunc) const {
+                                                     bool IsFunc,
+                                                     bool IsGlob) const {
   MCSymbolRefExpr::VariantKind VK =
-      IsFunc ? MCSymbolRefExpr::VK_WebAssembly_FUNCTION
+      IsFunc ? MCSymbolRefExpr::VK_WebAssembly_FUNCTION :
+      IsGlob ? MCSymbolRefExpr::VK_WebAssembly_GLOBAL
              : MCSymbolRefExpr::VK_None;
 
   const MCExpr *Expr = MCSymbolRefExpr::create(Sym, VK, Ctx);
 
   if (Offset != 0) {
     if (IsFunc)
       report_fatal_error("Function addresses with offsets not supported");
+    if (IsGlob)
+      report_fatal_error("Global indexes with offsets not supported");
     Expr =
         MCBinaryExpr::createAdd(Expr, MCConstantExpr::create(Offset, Ctx), Ctx);
   }
 
   return MCOperand::createExpr(Expr);
 }
 
 // Return the WebAssembly type associated with the given register class.
case MachineOperand::MO_FPImmediate: {
         MCOp = MCOperand::createFPImm(Imm->getValueAPF().convertToFloat());
       else if (Imm->getType()->isDoubleTy())
         MCOp = MCOperand::createFPImm(Imm->getValueAPF().convertToDouble());
       else
         llvm_unreachable("unknown floating point immediate type");
       break;
     }
     case MachineOperand::MO_GlobalAddress:
-      assert(MO.getTargetFlags() == 0 &&
+      assert(MO.getTargetFlags() == WebAssemblyII::MO_NO_FLAG &&
              "WebAssembly does not use target flags on GlobalAddresses");
       MCOp = LowerSymbolOperand(GetGlobalAddressSymbol(MO), MO.getOffset(),
-                                MO.getGlobal()->getValueType()->isFunctionTy());
+                                MO.getGlobal()->getValueType()->isFunctionTy(),
+                                false);
       break;
     case MachineOperand::MO_ExternalSymbol:
       // The target flag indicates whether this is a symbol for a
       // variable or a function.
-      assert((MO.getTargetFlags() & -2) == 0 &&
-             "WebAssembly uses only one target flag bit on ExternalSymbols");
+      assert((MO.getTargetFlags() & ~WebAssemblyII::MO_SYMBOL_MASK) == 0 &&
+             "WebAssembly uses only symbol flags on ExternalSymbols");
       MCOp = LowerSymbolOperand(GetExternalSymbolSymbol(MO), /*Offset=*/0,
-                                MO.getTargetFlags() & 1);
+          (MO.getTargetFlags() & WebAssemblyII::MO_SYMBOL_FUNCTION) != 0,
+          (MO.getTargetFlags() & WebAssemblyII::MO_SYMBOL_GLOBAL) != 0);
       break;
     }
 
     OutMI.addOperand(MCOp);
   }
 }

test/CodeGen/WebAssembly/byval.ll

 declare void @ext_byval_func_alignedstruct(%AlignedStruct* byval)
 declare void @ext_byval_func_bigarray(%BigArray* byval)
 declare void @ext_byval_func_empty(%EmptyStruct* byval)
 
 ; CHECK-LABEL: byval_arg
 define void @byval_arg(%SmallStruct* %ptr) {
  ; CHECK: .param i32
  ; Subtract 16 from SP (SP is 16-byte aligned)
- ; CHECK-NEXT: get_global $push[[L2:.+]]=, __stack_pointer
+ ; CHECK-NEXT: get_global $push[[L2:.+]]=, __stack_pointer@GLOBAL
  ; CHECK-NEXT: i32.const $push[[L3:.+]]=, 16
  ; CHECK-NEXT: i32.sub $push[[L11:.+]]=, $pop[[L2]], $pop[[L3]]
  ; Ensure SP is stored back before the call
  ; CHECK-NEXT: tee_local $push[[L10:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L10]]{{$}}
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L10]]{{$}}
  ; Copy the SmallStruct argument to the stack (SP+12, original SP-4)
  ; CHECK-NEXT: i32.load $push[[L0:.+]]=, 0($0)
  ; CHECK-NEXT: i32.store 12($[[SP]]), $pop[[L0]]
  ; Pass a pointer to the stack slot to the function
  ; CHECK-NEXT: i32.const $push[[L5:.+]]=, 12{{$}}
  ; CHECK-NEXT: i32.add $push[[ARG:.+]]=, $[[SP]], $pop[[L5]]{{$}}
  ; CHECK-NEXT: call ext_byval_func@FUNCTION, $pop[[ARG]]{{$}}
  call void @ext_byval_func(%SmallStruct* byval %ptr)
  ; Restore the stack
  ; CHECK-NEXT: i32.const $push[[L6:.+]]=, 16
  ; CHECK-NEXT: i32.add $push[[L8:.+]]=, $[[SP]], $pop[[L6]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L8]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L8]]
  ; CHECK-NEXT: return
  ret void
 }
 
 ; CHECK-LABEL: byval_arg_align8
 define void @byval_arg_align8(%SmallStruct* %ptr) {
  ; CHECK: .param i32
  ; Don't check the entire SP sequence, just enough to get the alignment.
  ; CHECK: i32.const $push[[L1:.+]]=, 16
  ; CHECK-NEXT: i32.sub $push[[L11:.+]]=, {{.+}}, $pop[[L1]]
  ; CHECK-NEXT: tee_local $push[[L10:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L10]]{{$}}
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L10]]{{$}}
  ; Copy the SmallStruct argument to the stack (SP+8, original SP-8)
  ; CHECK-NEXT: i32.load $push[[L0:.+]]=, 0($0){{$}}
  ; CHECK-NEXT: i32.store 8($[[SP]]), $pop[[L0]]{{$}}
  ; Pass a pointer to the stack slot to the function
  ; CHECK-NEXT: i32.const $push[[L5:.+]]=, 8{{$}}
  ; CHECK-NEXT: i32.add $push[[ARG:.+]]=, $[[SP]], $pop[[L5]]{{$}}
  ; CHECK-NEXT: call ext_byval_func_align8@FUNCTION, $pop[[ARG]]{{$}}
  call void @ext_byval_func_align8(%SmallStruct* byval align 8 %ptr)
  ret void
 }
 
 ; CHECK-LABEL: byval_arg_double
 define void @byval_arg_double(%AlignedStruct* %ptr) {
  ; CHECK: .param i32
  ; Subtract 16 from SP (SP is 16-byte aligned)
  ; CHECK: i32.const $push[[L1:.+]]=, 16
  ; CHECK-NEXT: i32.sub $push[[L14:.+]]=, {{.+}}, $pop[[L1]]
  ; CHECK-NEXT: tee_local $push[[L13:.+]]=, $[[SP:.+]]=, $pop[[L14]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L13]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L13]]
  ; Copy the AlignedStruct argument to the stack (SP+0, original SP-16)
  ; Just check the last load/store pair of the memcpy
  ; CHECK: i64.load $push[[L4:.+]]=, 0($0)
  ; CHECK-NEXT: i64.store 0($[[SP]]), $pop[[L4]]
  ; Pass a pointer to the stack slot to the function
  ; CHECK-NEXT: call ext_byval_func_alignedstruct@FUNCTION, $[[SP]]
  tail call void @ext_byval_func_alignedstruct(%AlignedStruct* byval %ptr)
  ret void

  ; CHECK: .param i32
  ; CHECK: call ext_func_empty@FUNCTION, $0
  call void @ext_func_empty(%EmptyStruct* %ptr)
  ret void
 }
 
 ; Call memcpy for "big" byvals.
 ; CHECK-LABEL: big_byval:
-; CHECK:      get_global $push[[L2:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L2:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: i32.const $push[[L3:.+]]=, 131072
 ; CHECK-NEXT: i32.sub $push[[L11:.+]]=, $pop[[L2]], $pop[[L3]]
 ; CHECK-NEXT: tee_local $push[[L10:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[L10]]{{$}}
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L10]]{{$}}
 ; CHECK-NEXT: i32.const $push[[L0:.+]]=, 131072
 ; CHECK-NEXT: i32.call       $push[[L11:.+]]=, memcpy@FUNCTION, $[[SP]], ${{.+}}, $pop{{.+}}
 ; CHECK-NEXT: tee_local      $push[[L9:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
 ; CHECK-NEXT: call           big_byval_callee@FUNCTION,
 %big = type [131072 x i8]
 declare void @big_byval_callee(%big* byval align 1)
 define void @big_byval(%big* byval align 1 %x) {
   call void @big_byval_callee(%big* byval align 1 %x)
   ret void
 }

test/CodeGen/WebAssembly/reg-stackify.ll

 bb10:                                             ; preds = %bb9, %bb
   ret void
 }
 
 ; Don't stackify a call past a __stack_pointer store.
 
 ; CHECK-LABEL: stackpointer_dependency:
 ; CHECK:      call {{.+}}, stackpointer_callee@FUNCTION,
-; CHECK-NEXT: set_global __stack_pointer,
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL,
 declare i32 @stackpointer_callee(i8* readnone, i8* readnone)
 declare i8* @llvm.frameaddress(i32)
 define i32 @stackpointer_dependency(i8* readnone) {
   %2 = tail call i8* @llvm.frameaddress(i32 0)
   %3 = tail call i32 @stackpointer_callee(i8* %0, i8* %2)
   ret i32 %3
 }
 

test/CodeGen/WebAssembly/stack-alignment.ll

 ; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt | FileCheck %s
 
 target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
 target triple = "wasm32-unknown-unknown-wasm"
 
 declare void @somefunc(i32*)
 
 ; CHECK-LABEL: underalign:
-; CHECK:      get_global $push[[L1:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L1:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: i32.const $push[[L2:.+]]=, 16
 ; CHECK-NEXT: i32.sub   $push[[L10:.+]]=, $pop[[L1]], $pop[[L2]]
 ; CHECK-NEXT: tee_local $push{{.+}}=, [[SP:.+]], $pop[[L10]]
 
 ; CHECK:      get_local $push[[L3:.+]]=, [[SP]]{{$}}
 ; CHECK:      i32.add   $push[[underaligned:.+]]=, $pop[[L3]], $pop{{.+}}
 ; CHECK-NEXT: call      somefunc@FUNCTION, $pop[[underaligned]]
 
 ; CHECK:      get_local $push[[M4:.+]]=, [[SP]]{{$}}
 ; CHECK:      i32.add   $push[[L5:.+]]=, $pop[[M4]], $pop{{.+}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[L5]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L5]]
 define void @underalign() {
 entry:
   %underaligned = alloca i32, align 8
   call void @somefunc(i32* %underaligned)
   ret void
 }
 
 ; CHECK-LABEL: overalign:
-; CHECK:      get_global $push[[L10:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L10:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: tee_local  $push[[L9:.+]]=, [[BP:.+]], $pop[[L10]]
 ; CHECK-NEXT: i32.const  $push[[L2:.+]]=, 32
 ; CHECK-NEXT: i32.sub    $push[[L8:.+]]=, $pop[[L9]], $pop[[L2]]
 ; CHECK-NEXT: i32.const  $push[[L3:.+]]=, -32
 ; CHECK-NEXT: i32.and    $push[[L7:.+]]=, $pop[[L8]], $pop[[L3]]
 ; CHECK-NEXT: tee_local  $push{{.+}}=, [[SP:.+]], $pop[[L7]]
 
 ; CHECK:      get_local  $push[[M5:.+]]=, [[SP]]{{$}}
 ; CHECK:      call       somefunc@FUNCTION, $pop[[M5]]{{$}}
 
 ; CHECK:      get_local  $push[[M6:.+]]=, [[BP]]{{$}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[M6]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[M6]]
 define void @overalign() {
 entry:
   %overaligned = alloca i32, align 32
   call void @somefunc(i32* %overaligned)
   ret void
 }
 
 ; CHECK-LABEL: over_and_normal_align:
-; CHECK:      get_global $push[[L14:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L14:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: tee_local  $push[[L13:.+]]=, [[BP:.+]], $pop[[L14]]
 ; CHECK:      i32.sub    $push[[L12:.+]]=, $pop[[L13]], $pop{{.+}}
 ; CHECK:      i32.and    $push[[L11:.+]]=, $pop[[L12]], $pop{{.+}}
 ; CHECK-NEXT: tee_local  $push{{.+}}=, [[SP:.+]], $pop[[L11]]
 
 ; CHECK:      get_local  $push[[M6:.+]]=, [[SP]]{{$}}
 ; CHECK:      i32.add    $push[[L6:.+]]=, $pop[[M6]], $pop{{.+}}
 ; CHECK-NEXT: call       somefunc@FUNCTION, $pop[[L6]]
 ; CHECK:      get_local  $push[[M7:.+]]=, [[SP]]{{$}}
 ; CHECK:      i32.add    $push[[L8:.+]]=, $pop[[M7]], $pop{{.+}}
 ; CHECK-NEXT: call       somefunc@FUNCTION, $pop[[L8]]
 
 ; CHECK:      get_local  $push[[L6:.+]]=, [[BP]]{{$}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[L6]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L6]]
 define void @over_and_normal_align() {
 entry:
   %over = alloca i32, align 32
   %normal = alloca i32
   call void @somefunc(i32* %over)
   call void @somefunc(i32* %normal)
   ret void
 }
 
 ; CHECK-LABEL: dynamic_overalign:
-; CHECK:      get_global $push[[L18:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L18:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: tee_local  $push[[L17:.+]]=, [[SP:.+]], $pop[[L18]]
 ; CHECK-NEXT: set_local  [[BP:.+]], $pop[[L17]]
 ; CHECK:      tee_local  $push{{.+}}=, [[SP_2:.+]], $pop{{.+}}
 
 ; CHECK:      get_local  $push[[M8:.+]]=, [[SP_2]]{{$}}
 ; CHECK:      call       somefunc@FUNCTION, $pop[[M8]]
 
 ; CHECK:      get_local  $push[[M9:.+]]=, [[BP]]{{$}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[M9]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[M9]]
 define void @dynamic_overalign(i32 %num) {
 entry:
   %dynamic = alloca i32, i32 %num, align 32
   call void @somefunc(i32* %dynamic)
   ret void
 }
 
 ; CHECK-LABEL: overalign_and_dynamic:
-; CHECK:      get_global $push[[L21:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L21:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: tee_local  $push[[L20:.+]]=, [[BP:.+]], $pop[[L21]]
 ; CHECK:      i32.sub    $push[[L19:.+]]=, $pop[[L20]], $pop{{.+}}
 ; CHECK:      i32.and    $push[[L18:.+]]=, $pop[[L19]], $pop{{.+}}
 ; CHECK:      tee_local  $push{{.+}}=, [[FP:.+]], $pop[[L18]]
 ; CHECK:      get_local  $push[[M10:.+]]=, [[FP]]{{$}}
 ; CHECK:      i32.sub    $push[[L16:.+]]=, $pop[[M10]], $pop{{.+}}
 ; CHECK-NEXT: tee_local  $push{{.+}}=, [[SP:.+]], $pop[[L16]]
 
 ; CHECK:      get_local  $push[[over:.+]]=, [[FP]]
 ; CHECK-NEXT: call       somefunc@FUNCTION, $pop[[over]]
 ; CHECK:      get_local  $push[[another:.+]]=, [[SP]]
 ; CHECK-NEXT: call       somefunc@FUNCTION, $pop[[another]]
 
 ; CHECK:      get_local  $push[[M11:.+]]=, [[BP]]{{$}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[M11]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[M11]]
 define void @overalign_and_dynamic(i32 %num) {
 entry:
   %over = alloca i32, align 32
   %dynamic = alloca i32, i32 %num
   call void @somefunc(i32* %over)
   call void @somefunc(i32* %dynamic)
   ret void
 }
 
 ; CHECK-LABEL: overalign_static_and_dynamic:
-; CHECK:      get_global $push[[L26:.+]]=, __stack_pointer{{$}}
+; CHECK:      get_global $push[[L26:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: tee_local  $push[[L25:.+]]=, [[BP:.+]], $pop[[L26]]
 ; CHECK:      i32.sub    $push[[L24:.+]]=, $pop[[L25]], $pop{{.+}}
 ; CHECK:      i32.and    $push[[L23:.+]]=, $pop[[L24]], $pop{{.+}}
 ; CHECK:      tee_local  $push{{.+}}=, [[FP:.+]], $pop[[L23]]
 ; CHECK:      get_local  $push[[M12:.+]]=, [[FP]]{{$}}
 ; CHECK:      i32.sub    $push[[L21:.+]]=, $pop[[M12]], $pop{{.+}}
 ; CHECK-NEXT: tee_local  $push{{.+}}=, [[SP:.+]], $pop[[L21]]
 
 ; CHECK:      get_local  $push[[L19:.+]]=, [[FP]]
 ; CHECK:      tee_local  $push[[L18:.+]]=, [[FP_2:.+]], $pop[[L19]]
 ; CHECK:      i32.add    $push[[over:.+]]=, $pop[[L18]], $pop{{.+}}
 ; CHECK-NEXT: call       somefunc@FUNCTION, $pop[[over]]
 ; CHECK:      get_local  $push[[M12:.+]]=, [[SP]]
 ; CHECK:      call       somefunc@FUNCTION, $pop[[M12]]
 ; CHECK:      get_local  $push[[M13:.+]]=, [[FP_2]]
 ; CHECK:      i32.add    $push[[static:.+]]=, $pop[[M13]], $pop{{.+}}
 ; CHECK-NEXT: call       somefunc@FUNCTION, $pop[[static]]
 
 ; CHECK:      get_local  $push[[M14:.+]]=, [[BP]]{{$}}
-; CHECK-NEXT: set_global __stack_pointer, $pop[[M14]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[M14]]
 define void @overalign_static_and_dynamic(i32 %num) {
 entry:
   %over = alloca i32, align 32
   %dynamic = alloca i32, i32 %num
   %static = alloca i32
   call void @somefunc(i32* %over)
   call void @somefunc(i32* %dynamic)
   call void @somefunc(i32* %static)
   ret void
 }

test/CodeGen/WebAssembly/userstack.ll

 ; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt | FileCheck %s
 
 target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
 target triple = "wasm32-unknown-unknown-wasm"
 
 declare void @ext_func(i64* %ptr)
 declare void @ext_func_i32(i32* %ptr)
 
 ; CHECK-LABEL: alloca32:
 ; Check that there is an extra local for the stack pointer.
 ; CHECK: .local i32{{$}}
 define void @alloca32() noredzone {
- ; CHECK-NEXT: get_global $push[[L2:.+]]=, __stack_pointer{{$}}
+ ; CHECK-NEXT: get_global $push[[L2:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: i32.const $push[[L3:.+]]=, 16
  ; CHECK-NEXT: i32.sub $push[[L9:.+]]=, $pop[[L2]], $pop[[L3]]
  ; CHECK-NEXT: tee_local $push[[L8:.+]]=, [[SP:.+]], $pop[[L9]]{{$}}
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L8]]{{$}}
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L8]]{{$}}
  %retval = alloca i32
  ; CHECK: get_local $push[[L4:.+]]=, [[SP]]{{$}}
  ; CHECK: i32.const $push[[L0:.+]]=, 0
  ; CHECK: i32.store 12($pop[[L4]]), $pop[[L0]]
  store i32 0, i32* %retval
  ; CHECK: get_local $push[[L6:.+]]=, [[SP]]{{$}}
  ; CHECK-NEXT: i32.const $push[[L5:.+]]=, 16
  ; CHECK-NEXT: i32.add $push[[L7:.+]]=, $pop[[L6]], $pop[[L5]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L7]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L7]]
  ret void
 }
 
 ; CHECK-LABEL: alloca3264:
 ; CHECK: .local i32{{$}}
 define void @alloca3264() {
- ; CHECK: get_global $push[[L3:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L3:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: i32.const $push[[L4:.+]]=, 16
  ; CHECK-NEXT: i32.sub $push[[L6:.+]]=, $pop[[L3]], $pop[[L4]]
  ; CHECK-NEXT: tee_local $push[[L5:.+]]=, [[SP:.+]], $pop[[L6]]
  %r1 = alloca i32
  %r2 = alloca double
  store i32 0, i32* %r1
  store double 0.0, double* %r2
  ; CHECK-NEXT: i64.const $push[[L1:.+]]=, 0
  ; CHECK-NEXT: i64.store 0($pop[[L5]]), $pop[[L1]]
  ; CHECK-NEXT: get_local $push[[L2:.+]]=, [[SP]]{{$}}
  ; CHECK-NEXT: i32.const $push[[L0:.+]]=, 0
  ; CHECK-NEXT: i32.store 12($pop[[L2]]), $pop[[L0]]
  ; CHECK-NEXT: return
  ret void
 }
 
 ; CHECK-LABEL: allocarray:
 ; CHECK: .local i32{{$}}
 define void @allocarray() {
- ; CHECK-NEXT: get_global $push[[L4:.+]]=, __stack_pointer{{$}}
+ ; CHECK-NEXT: get_global $push[[L4:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: i32.const $push[[L5:.+]]=, 144{{$}}
  ; CHECK-NEXT: i32.sub $push[[L12:.+]]=, $pop[[L4]], $pop[[L5]]
  ; CHECK-NEXT: tee_local $push[[L11:.+]]=, 0, $pop[[L12]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L11]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L11]]
  %r = alloca [33 x i32]
 
  ; CHECK:      i32.const $push{{.+}}=, 24
  ; CHECK-NEXT: i32.add $push[[L3:.+]]=, $pop{{.+}}, $pop{{.+}}
  ; CHECK-NEXT: i32.const $push[[L1:.+]]=, 1{{$}}
  ; CHECK-NEXT: i32.store 0($pop[[L3]]), $pop[[L1]]{{$}}
  ; CHECK-NEXT: get_local $push[[L4:.+]]=, 0{{$}}
  ; CHECK-NEXT: i32.const $push[[L10:.+]]=, 1{{$}}
  ; CHECK-NEXT: i32.store 12($pop[[L4]]), $pop[[L10]]{{$}}
  %p = getelementptr [33 x i32], [33 x i32]* %r, i32 0, i32 0
  store i32 1, i32* %p
  %p2 = getelementptr [33 x i32], [33 x i32]* %r, i32 0, i32 3
  store i32 1, i32* %p2
 
  ; CHECK-NEXT: get_local $push[[L2:.+]]=, [[SP]]{{$}}
  ; CHECK-NEXT: i32.const $push[[L7:.+]]=, 144
  ; CHECK-NEXT: i32.add $push[[L8:.+]]=, $pop[[L2]], $pop[[L7]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L8]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L8]]
  ret void
 }
 
 ; CHECK-LABEL: non_mem_use
 define void @non_mem_use(i8** %addr) {
  ; CHECK: i32.const $push[[L2:.+]]=, 48
  ; CHECK-NEXT: i32.sub $push[[L12:.+]]=, {{.+}}, $pop[[L2]]
  ; CHECK-NEXT: tee_local $push[[L11:.+]]=, [[SP:.+]], $pop[[L12]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L11]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L11]]
  %buf = alloca [27 x i8], align 16
  %r = alloca i64
  %r2 = alloca i64
  ; %r is at SP+8
  ; CHECK: get_local $push[[L3:.+]]=, [[SP]]
  ; CHECK: i32.const $push[[OFF:.+]]=, 8
  ; CHECK-NEXT: i32.add $push[[ARG1:.+]]=, $pop[[L3]], $pop[[OFF]]
  ; CHECK-NEXT: call ext_func@FUNCTION, $pop[[ARG1]]

  %gep = getelementptr inbounds [27 x i8], [27 x i8]* %buf, i32 0, i32 0
  store i8* %gep, i8** %addr
  ret void
 }
 
 ; CHECK-LABEL: allocarray_inbounds:
 ; CHECK: .local i32{{$}}
 define void @allocarray_inbounds() {
- ; CHECK: get_global $push[[L3:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L3:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: i32.const $push[[L4:.+]]=, 32{{$}}
  ; CHECK-NEXT: i32.sub $push[[L11:.+]]=, $pop[[L3]], $pop[[L4]]
  ; CHECK-NEXT: tee_local $push[[L10:.+]]=, [[SP:.+]], $pop[[L11]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L10]]{{$}}
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L10]]{{$}}
  %r = alloca [5 x i32]
  ; CHECK: i32.const $push[[L3:.+]]=, 1
  ; CHECK-DAG: i32.store 24(${{.+}}), $pop[[L3]]
  %p = getelementptr inbounds [5 x i32], [5 x i32]* %r, i32 0, i32 0
  store i32 1, i32* %p
  ; This store should have both the GEP and the FI folded into it.
  ; CHECK-DAG: i32.store 12(${{.+}}), $pop
  %p2 = getelementptr inbounds [5 x i32], [5 x i32]* %r, i32 0, i32 3
  store i32 1, i32* %p2
  call void @ext_func(i64* null);
  ; CHECK: call ext_func
  ; CHECK: i32.const $push[[L5:.+]]=, 32{{$}}
  ; CHECK-NEXT: i32.add $push[[L7:.+]]=, ${{.+}}, $pop[[L5]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L7]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L7]]
  ret void
 }
 
 ; CHECK-LABEL: dynamic_alloca:
 define void @dynamic_alloca(i32 %alloc) {
- ; CHECK: get_global $push[[L13:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L13:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: tee_local $push[[L12:.+]]=, [[SP:.+]], $pop[[L13]]{{$}}
  ; Target independent codegen bumps the stack pointer.
  ; CHECK: i32.sub
  ; Check that SP is written back to memory after decrement
- ; CHECK: set_global __stack_pointer, 
+ ; CHECK: set_global __stack_pointer@GLOBAL,
  %r = alloca i32, i32 %alloc
  ; Target-independent codegen also calculates the store addr
  ; CHECK: call ext_func_i32@FUNCTION
  call void @ext_func_i32(i32* %r)
- ; CHECK: set_global __stack_pointer, $pop{{.+}}
+ ; CHECK: set_global __stack_pointer@GLOBAL, $pop{{.+}}
  ret void
 }
 
 ; CHECK-LABEL: dynamic_alloca_redzone:
 define void @dynamic_alloca_redzone(i32 %alloc) {
- ; CHECK: get_global $push[[L13:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L13:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: tee_local $push[[L12:.+]]=, [[SP:.+]], $pop[[L13]]{{$}}
  ; Target independent codegen bumps the stack pointer
  ; CHECK: i32.sub
  %r = alloca i32, i32 %alloc
  ; CHECK-NEXT: tee_local       $push[[L8:.+]]=, {{.+}}, $pop
  ; CHECK: get_local $push[[L7:.+]]=, 0{{$}}
  ; CHECK-NEXT: i32.const       $push[[L6:.+]]=, 0{{$}}
  ; CHECK-NEXT: i32.store       0($pop[[L7]]), $pop[[L6]]{{$}}
  store i32 0, i32* %r
  ; CHECK-NEXT: return
  ret void
 }
 
 ; CHECK-LABEL: dynamic_static_alloca:
 define void @dynamic_static_alloca(i32 %alloc) noredzone {
  ; Decrement SP in the prolog by the static amount and writeback to memory.
- ; CHECK: get_global $push[[L11:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L11:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: i32.const $push[[L12:.+]]=, 16
  ; CHECK-NEXT: i32.sub $push[[L23:.+]]=, $pop[[L11]], $pop[[L12]]
  ; CHECK-NEXT: tee_local $push[[L22:.+]]=, [[SP:.+]], $pop[[L23]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L22]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L22]]
 
  ; Alloc and write to a static alloca
  ; CHECK: get_local $push[[L21:.+]]=, [[SP:.+]]
  ; CHECK-NEXT: tee_local $push[[pushedFP:.+]]=, [[FP:.+]], $pop[[L21]]
  ; CHECK-NEXT: i32.const $push[[L0:.+]]=, 101
  ; CHECK-NEXT: i32.store [[static_offset:.+]]($pop[[pushedFP]]), $pop[[L0]]
  %static = alloca i32
  store volatile i32 101, i32* %static
 
  ; Decrement SP in the body by the dynamic amount.
  ; CHECK: i32.sub
  ; CHECK: tee_local $push[[L16:.+]]=, [[dynamic_local:.+]], $pop{{.+}}
  ; CHECK: tee_local $push[[L15:.+]]=, [[other:.+]], $pop[[L16]]{{$}}
- ; CHECK: set_global __stack_pointer, $pop[[L15]]{{$}}
+ ; CHECK: set_global __stack_pointer@GLOBAL, $pop[[L15]]{{$}}
  %dynamic = alloca i32, i32 %alloc
 
  ; Ensure we don't modify the frame pointer after assigning it.
  ; CHECK-NOT: $[[FP]]=
 
  ; Ensure the static address doesn't change after modifying the stack pointer.
  ; CHECK: get_local $push[[L17:.+]]=, [[FP]]
  ; CHECK: i32.const $push[[L7:.+]]=, 102

  store volatile i32 104, i32* %static
  store volatile i32 105, i32* %dynamic
  store volatile i32 106, i32* %dynamic.2
 
  ; Writeback to memory.
  ; CHECK: get_local $push[[L24:.+]]=, [[FP]]{{$}}
  ; CHECK: i32.const $push[[L18:.+]]=, 16
  ; CHECK-NEXT: i32.add $push[[L19:.+]]=, $pop[[L24]], $pop[[L18]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L19]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L19]]
  ret void
 }
 
 declare i8* @llvm.stacksave()
 declare void @llvm.stackrestore(i8*)
 
 ; CHECK-LABEL: llvm_stack_builtins:
 define void @llvm_stack_builtins(i32 %alloc) noredzone {
- ; CHECK: get_global $push[[L11:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L11:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: tee_local $push[[L10:.+]]=, {{.+}}, $pop[[L11]]
  ; CHECK-NEXT: set_local [[STACK:.+]], $pop[[L10]]
  %stack = call i8* @llvm.stacksave()
 
  ; Ensure we don't reassign the stacksave local
  ; CHECK-NOT: set_local [[STACK]],
  %dynamic = alloca i32, i32 %alloc
 
  ; CHECK: get_local $push[[L12:.+]]=, [[STACK]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L12]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L12]]
  call void @llvm.stackrestore(i8* %stack)
 
  ret void
 }
 
 ; Not actually using the alloca'd variables exposed an issue with register
 ; stackification, where copying the stack pointer into the frame pointer was
 ; moved after the stack pointer was updated for the dynamic alloca.
 ; CHECK-LABEL: dynamic_alloca_nouse:
 define void @dynamic_alloca_nouse(i32 %alloc) noredzone {
- ; CHECK: get_global $push[[L11:.+]]=, __stack_pointer{{$}}
+ ; CHECK: get_global $push[[L11:.+]]=, __stack_pointer@GLOBAL{{$}}
  ; CHECK-NEXT: tee_local $push[[L10:.+]]=, {{.+}}, $pop[[L11]]
  ; CHECK-NEXT: set_local [[FP:.+]], $pop[[L10]]
  %dynamic = alloca i32, i32 %alloc
 
  ; CHECK-NOT: set_local [[FP]],
 
  ; CHECK: get_local $push[[L12:.+]]=, [[FP]]
- ; CHECK-NEXT: set_global __stack_pointer, $pop[[L12]]
+ ; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L12]]
  ret void
 }
 
 ; The use of the alloca in a phi causes a CopyToReg DAG node to be generated,
 ; which has to have special handling because CopyToReg can't have a FI operand
 ; CHECK-LABEL: copytoreg_fi:
 define void @copytoreg_fi(i1 %cond, i32* %b) {
 entry:

  ret void
 }
 
 declare void @use_i8_star(i8*)
 declare i8* @llvm.frameaddress(i32)
 
 ; Test __builtin_frame_address(0).
 ; CHECK-LABEL: frameaddress_0:
-; CHECK: get_global $push[[L3:.+]]=, __stack_pointer{{$}}
+; CHECK: get_global $push[[L3:.+]]=, __stack_pointer@GLOBAL{{$}}
 ; CHECK-NEXT: tee_local $push[[L2:.+]]=, [[FP:.+]], $pop[[L3]]{{$}}
 ; CHECK-NEXT: call use_i8_star@FUNCTION, $pop[[L2]]
 ; CHECK-NEXT: get_local $push[[L5:.+]]=, [[FP]]
-; CHECK-NEXT: set_global __stack_pointer, $pop[[L5]]
+; CHECK-NEXT: set_global __stack_pointer@GLOBAL, $pop[[L5]]
 define void @frameaddress_0() {
   %t = call i8* @llvm.frameaddress(i32 0)
   call void @use_i8_star(i8* %t)
   ret void
 }
 
 ; Test __builtin_frame_address(1).
 
 ; CHECK-LABEL: frameaddress_1:
 ; CHECK-NEXT: i32.const $push0=, 0{{$}}
 ; CHECK-NEXT: call use_i8_star@FUNCTION, $pop0{{$}}
 ; CHECK-NEXT: return{{$}}
 define void @frameaddress_1() {
   %t = call i8* @llvm.frameaddress(i32 1)
   call void @use_i8_star(i8* %t)
   ret void
 }
 
 ; Test a stack address passed to an inline asm.
 ; CHECK-LABEL: inline_asm:
-; CHECK:       get_global {{.+}}, __stack_pointer{{$}}
+; CHECK:       get_global {{.+}}, __stack_pointer@GLOBAL{{$}}
 ; CHECK:       #APP
 ; CHECK-NEXT:  # %{{[0-9]+}}{{$}}
 ; CHECK-NEXT:  #NO_APP
 define void @inline_asm() {
   %tmp = alloca i8
   call void asm sideeffect "# %0", "r"(i8* %tmp)
   ret void
 }
 
 ; TODO: test over-aligned alloca