[WebAssembly] Use 64-bit table when targeting wasm64 (llvm#92042)

See WebAssembly/memory64#51

Author: sbc100

lld/test/wasm/shared64.s

@@ -154,6 +154,7 @@ get_local_func_address:
 # CHECK-NEXT:           Index:           0
 # CHECK-NEXT:           ElemType:        FUNCREF
 # CHECK-NEXT:           Limits:
+# CHECK-NEXT:             Flags:         [ IS_64 ]
 # CHECK-NEXT:             Minimum:         0x2
 # CHECK-NEXT:       - Module:          env
 # CHECK-NEXT:         Field:           __stack_pointer
@@ -170,11 +171,6 @@ get_local_func_address:
 # CHECK-NEXT:         Kind:            GLOBAL
 # CHECK-NEXT:         GlobalType:      I64
 # CHECK-NEXT:         GlobalMutable:   false
-# CHECK-NEXT:       - Module:          env
-# CHECK-NEXT:         Field:           __table_base32
-# CHECK-NEXT:         Kind:            GLOBAL
-# CHECK-NEXT:         GlobalType:      I32
-# CHECK-NEXT:         GlobalMutable:   false
 # CHECK-NEXT:       - Module:          GOT.mem
 # CHECK-NEXT:         Field:           indirect_func
 # CHECK-NEXT:         Kind:            GLOBAL
@@ -209,7 +205,7 @@ get_local_func_address:
 # CHECK-NEXT:     Segments:
 # CHECK-NEXT:       - Offset:
 # CHECK-NEXT:           Opcode:          GLOBAL_GET
-# CHECK-NEXT:           Index:           3
+# CHECK-NEXT:           Index:           2
 # CHECK-NEXT:         Functions:       [ 3, 2 ]
 
 # check the generated code in __wasm_call_ctors and __wasm_apply_data_relocs functions
@@ -223,7 +219,7 @@ get_local_func_address:
 # DIS-NEXT:                 i64.const       4
 # DIS-NEXT:                 global.get      1
 # DIS-NEXT:                 i64.add
-# DIS-NEXT:                 global.get      5
+# DIS-NEXT:                 global.get      4
 # DIS-NEXT:                 i64.store       0:p2align=2
 # DIS-NEXT:                 i64.const       12
 # DIS-NEXT:                 global.get      1
@@ -242,12 +238,12 @@ get_local_func_address:
 # DIS-NEXT:                 i64.const       24
 # DIS-NEXT:                 global.get      1
 # DIS-NEXT:                 i64.add
-# DIS-NEXT:                 global.get      6
+# DIS-NEXT:                 global.get      5
 # DIS-NEXT:                 i64.store       0:p2align=2
 # DIS-NEXT:                 i64.const       32
 # DIS-NEXT:                 global.get      1
 # DIS-NEXT:                 i64.add
-# DIS-NEXT:                 global.get      7
+# DIS-NEXT:                 global.get      6
 # DIS-NEXT:                 i32.const       4
 # DIS-NEXT:                 i32.add
 # DIS-NEXT:                 i32.store       0

lld/wasm/Driver.cpp

@@ -870,13 +870,6 @@ static void createSyntheticSymbols() {
     WasmSym::tableBase = createUndefinedGlobal("__table_base", globalType);
     WasmSym::memoryBase->markLive();
     WasmSym::tableBase->markLive();
-    if (is64) {
-      WasmSym::tableBase32 =
-          createUndefinedGlobal("__table_base32", &globalTypeI32);
-      WasmSym::tableBase32->markLive();
-    } else {
-      WasmSym::tableBase32 = nullptr;
-    }
   } else {
     // For non-PIC code
     WasmSym::stackPointer = createGlobalVariable("__stack_pointer", true);
@@ -923,9 +916,6 @@ static void createOptionalSymbols() {
     WasmSym::heapEnd = symtab->addOptionalDataSymbol("__heap_end");
     WasmSym::definedMemoryBase = symtab->addOptionalDataSymbol("__memory_base");
     WasmSym::definedTableBase = symtab->addOptionalDataSymbol("__table_base");
-    if (config->is64.value_or(false))
-      WasmSym::definedTableBase32 =
-          symtab->addOptionalDataSymbol("__table_base32");
   }
 
   // For non-shared memory programs we still need to define __tls_base since we

lld/wasm/Symbols.cpp

@@ -96,8 +96,6 @@ GlobalSymbol *WasmSym::tlsSize;
 GlobalSymbol *WasmSym::tlsAlign;
 UndefinedGlobal *WasmSym::tableBase;
 DefinedData *WasmSym::definedTableBase;
-UndefinedGlobal *WasmSym::tableBase32;
-DefinedData *WasmSym::definedTableBase32;
 UndefinedGlobal *WasmSym::memoryBase;
 DefinedData *WasmSym::definedMemoryBase;
 TableSymbol *WasmSym::indirectFunctionTable;

lld/wasm/Symbols.h

@@ -603,11 +603,6 @@ struct WasmSym {
   // Used in PIC code for offset of indirect function table
   static UndefinedGlobal *tableBase;
   static DefinedData *definedTableBase;
-  // 32-bit copy in wasm64 to work around init expr limitations.
-  // These can potentially be removed again once we have
-  // https://github.com/WebAssembly/extended-const 
-  static UndefinedGlobal *tableBase32;
-  static DefinedData *definedTableBase32;
 
   // __memory_base
   // Used in PIC code for offset of global data

lld/wasm/SyntheticSections.cpp

@@ -584,12 +584,10 @@ void ElemSection::writeBody() {
   initExpr.Extended = false;
   if (ctx.isPic) {
     initExpr.Inst.Opcode = WASM_OPCODE_GLOBAL_GET;
-    initExpr.Inst.Value.Global =
-        (config->is64.value_or(false) ? WasmSym::tableBase32
-                                      : WasmSym::tableBase)
-            ->getGlobalIndex();
+    initExpr.Inst.Value.Global = WasmSym::tableBase->getGlobalIndex();
   } else {
-    initExpr.Inst.Opcode = WASM_OPCODE_I32_CONST;
+    bool is64 = config->is64.value_or(false);
+    initExpr.Inst.Opcode = is64 ? WASM_OPCODE_I64_CONST : WASM_OPCODE_I32_CONST;
     initExpr.Inst.Value.Int32 = config->tableBase;
   }
   writeInitExpr(os, initExpr);

lld/wasm/Writer.cpp

@@ -939,6 +939,8 @@ static void finalizeIndirectFunctionTable() {
     limits.Flags |= WASM_LIMITS_FLAG_HAS_MAX;
     limits.Maximum = limits.Minimum;
   }
+  if (config->is64.value_or(false))
+    limits.Flags |= WASM_LIMITS_FLAG_IS_64;
   WasmSym::indirectFunctionTable->setLimits(limits);
 }
 
@@ -1691,12 +1693,8 @@ void Writer::createSyntheticSectionsPostLayout() {
 void Writer::run() {
   // For PIC code the table base is assigned dynamically by the loader.
   // For non-PIC, we start at 1 so that accessing table index 0 always traps.
-  if (!ctx.isPic) {
-    if (WasmSym::definedTableBase)
-      WasmSym::definedTableBase->setVA(config->tableBase);
-    if (WasmSym::definedTableBase32)
-      WasmSym::definedTableBase32->setVA(config->tableBase);
-  }
+  if (!ctx.isPic && WasmSym::definedTableBase)
+    WasmSym::definedTableBase->setVA(config->tableBase);
 
   log("-- createOutputSegments");
   createOutputSegments();

llvm/lib/Target/WebAssembly/WebAssemblyFastISel.cpp

@@ -885,18 +885,6 @@ bool WebAssemblyFastISel::selectCall(const Instruction *I) {
       Table->setNoStrip();
       MIB.addImm(0);
     }
-    // See if we must truncate the function pointer.
-    // CALL_INDIRECT takes an i32, but in wasm64 we represent function pointers
-    // as 64-bit for uniformity with other pointer types.
-    // See also: WebAssemblyISelLowering.cpp: LowerCallResults
-    if (Subtarget->hasAddr64()) {
-      auto Wrap = BuildMI(*FuncInfo.MBB, std::prev(FuncInfo.InsertPt), MIMD,
-                          TII.get(WebAssembly::I32_WRAP_I64));
-      Register Reg32 = createResultReg(&WebAssembly::I32RegClass);
-      Wrap.addReg(Reg32, RegState::Define);
-      Wrap.addReg(CalleeReg);
-      CalleeReg = Reg32;
-    }
   }
 
   for (unsigned ArgReg : Args)

llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

@@ -576,20 +576,6 @@ LowerCallResults(MachineInstr &CallResults, DebugLoc DL, MachineBasicBlock *BB,
   const MCInstrDesc &MCID = TII.get(CallOp);
   MachineInstrBuilder MIB(MF, MF.CreateMachineInstr(MCID, DL));
 
-  // See if we must truncate the function pointer.
-  // CALL_INDIRECT takes an i32, but in wasm64 we represent function pointers
-  // as 64-bit for uniformity with other pointer types.
-  // See also: WebAssemblyFastISel::selectCall
-  if (IsIndirect && MF.getSubtarget<WebAssemblySubtarget>().hasAddr64()) {
-    Register Reg32 =
-        MF.getRegInfo().createVirtualRegister(&WebAssembly::I32RegClass);
-    auto &FnPtr = CallParams.getOperand(0);
-    BuildMI(*BB, CallResults.getIterator(), DL,
-            TII.get(WebAssembly::I32_WRAP_I64), Reg32)
-        .addReg(FnPtr.getReg());
-    FnPtr.setReg(Reg32);
-  }
-
   // Move the function pointer to the end of the arguments for indirect calls
   if (IsIndirect) {
     auto FnPtr = CallParams.getOperand(0);

llvm/test/CodeGen/WebAssembly/fast-isel-call-indirect64.ll

@@ -34,7 +34,6 @@ entry:
 ; CHECK:      .functype foo (i64) -> ()
 ; CHECK-NEXT: i32.const 1
 ; CHECK-NEXT: local.get 0
-; CHECK-NEXT: i32.wrap_i64
 ; CHECK-NEXT: call_indirect (i32) -> ()
 ; REF:        call_indirect __indirect_function_table, (i32) -> ()
 
@@ -53,10 +52,10 @@ entry:
 ; YAML:      - Type:   CODE
 ; YAML:      - Type:   R_WASM_TABLE_INDEX_SLEB64
 ; YAML-NEXT:   Index:  0
-; YAML-NEXT:   Offset: 0x16
+; YAML-NEXT:   Offset: 0x15
 ; YAML:      - Type:   R_WASM_TABLE_INDEX_SLEB64
 ; YAML-NEXT:   Index:  0
-; YAML-NEXT:   Offset: 0x29
+; YAML-NEXT:   Offset: 0x28
 
 ; YAML:      - Type:   DATA
 ; YAML:      - Type:   R_WASM_TABLE_INDEX_I64