[WebAssembly] 64-bit memory limits

[aardappel]

Summary

Making sure we're supporting writing out 64-bit limits structures in object files + related code cleanup.

[aardappel]

Not much happening here, other than updating the memory limits writing. Probably needs an LLD specific test, @sbc100 ?

[sbc100]

Nice!

[sbc100]

Do we have wasm-ld tests for this stuff yet? If so, I would expect to see a test update in lld/test/wasm too.

[aardappel]

@sbc100 any suggestion for which test would be best to add a 64-bit version of? Most important thing is to ensure LLD is outputting 64-bit memory limits

[sbc100]

Maybe add a wasm64 section to lld/test/wasm/data-layout.ll? Shame i didn't convert that to .s format already ..

[aardappel]

Added part of LLD data-layout.ll test

[sbc100]

lld/test/wasm/data-layout.ll

6

I would give each object a unique name e.g. hello64.o ?

Than you can avoid rebuilding it below.

[aardappel]

Test uses 32/64-bit specific .o files

[aardappel]

Already landed.

lld/test/wasm/data-layout.ll

 ; RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown %p/Inputs/hello.s -o %t.hello.o
-; RUN: llc -filetype=obj %s -o %t.o
-
-target triple = "wasm32-unknown-unknown"
+; RUN: llc -mtriple=wasm32-unknown-unknown -filetype=obj %s -o %t.o
+; RUN: wasm-ld -m wasm32 -no-gc-sections --export=__data_end --export=__heap_base --allow-undefined --no-entry -o %t.wasm %t.o %t.hello.o
+; RUN: obj2yaml %t.wasm | FileCheck --check-prefixes CHECK,CHK32 %s
+
+; RUN: llvm-mc -filetype=obj -triple=wasm64-unknown-unknown %p/Inputs/hello.s -o %t.hello.o
+; RUN: llc -mtriple=wasm64-unknown-unknown -filetype=obj %s -o %t.o
+; RUN: wasm-ld -m wasm64 -no-gc-sections --export=__data_end --export=__heap_base --allow-undefined --no-entry -o %t.wasm %t.o %t.hello.o
+; RUN: obj2yaml %t.wasm | FileCheck --check-prefixes CHECK,CHK64 %s
 
 @foo = hidden global i32 1, align 4
 @aligned_bar = hidden global i32 3, align 16
 
 @hello_str = external global i8*
 @external_ref = global i8** @hello_str, align 8
 
 %struct.s = type { i32, i32 }
 @local_struct = hidden global %struct.s zeroinitializer, align 4
 @local_struct_internal_ptr = hidden local_unnamed_addr global i32* getelementptr inbounds (%struct.s, %struct.s* @local_struct, i32 0, i32 1), align 4
 
-; RUN: wasm-ld -no-gc-sections --export=__data_end --export=__heap_base --allow-undefined --no-entry -o %t.wasm %t.o %t.hello.o
-; RUN: obj2yaml %t.wasm | FileCheck %s
-
 ; CHECK:        - Type:            MEMORY
 ; CHECK-NEXT:     Memories:
-; CHECK-NEXT:       - Initial:         0x00000002
+; CHK32-NEXT:       - Initial:         0x00000002
+; CHK64-NEXT:       - Flags:           [ IS_64 ]
+; CHK64-NEXT:         Initial:         0x00000002
 ; CHECK-NEXT:   - Type:            GLOBAL
 ; CHECK-NEXT:     Globals:
 ; CHECK-NEXT:       - Index:           0
-; CHECK-NEXT:         Type:            I32
+; CHK32-NEXT:         Type:            I32
+; CHK64-NEXT:         Type:            I64
 ; CHECK-NEXT:         Mutable:         true
 ; CHECK-NEXT:         InitExpr:
-; CHECK-NEXT:           Opcode:          I32_CONST
+; CHK32-NEXT:           Opcode:          I32_CONST
+; CHK64-NEXT:           Opcode:          I64_CONST
 ; CHECK-NEXT:           Value:           66624
 ; CHECK-NEXT:       - Index:           1
 ; CHECK-NEXT:         Type:            I32
 ; CHECK-NEXT:         Mutable:         false
 ; CHECK-NEXT:         InitExpr:
 ; CHECK-NEXT:           Opcode:          I32_CONST
-; CHECK-NEXT:           Value:           1080
+; CHK32-NEXT:           Value:           1080
+; CHK64-NEXT:           Value:           1088
 ; CHECK-NEXT:       - Index:           2
 ; CHECK-NEXT:         Type:            I32
 ; CHECK-NEXT:         Mutable:         false
 ; CHECK-NEXT:         InitExpr:
 ; CHECK-NEXT:           Opcode:          I32_CONST
 ; CHECK-NEXT:           Value:           66624
 
 ; CHECK:        - Type:            DATA
 ; CHECK-NEXT:     Segments:
 ; CHECK-NEXT:       - SectionOffset:   7
 ; CHECK-NEXT:         InitFlags:       0
 ; CHECK-NEXT:         Offset:
 ; CHECK-NEXT:           Opcode:          I32_CONST
 ; CHECK-NEXT:           Value:           1024
 ; CHECK-NEXT:         Content:         68656C6C6F0A00
 ; CHECK-NEXT:       - SectionOffset:   20
 ; CHECK-NEXT:         InitFlags:       0
 ; CHECK-NEXT:         Offset:
 ; CHECK-NEXT:           Opcode:          I32_CONST
 ; CHECK-NEXT:           Value:           1040
-; CHECK-NEXT:         Content:         '0100000000000000000000000000000003000000000000000004000034040000'
-; CHECK-NEXT:    - Type:            CUSTOM
 
 
+; RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown %p/Inputs/hello.s -o %t.hello.o
+; RUN: llc -mtriple=wasm32-unknown-unknown -filetype=obj %s -o %t.o
+
 ; RUN: wasm-ld -no-gc-sections --allow-undefined --no-entry \
 ; RUN:     --initial-memory=131072 --max-memory=131072 -o %t_max.wasm %t.o \
 ; RUN:     %t.hello.o
 ; RUN: obj2yaml %t_max.wasm | FileCheck %s -check-prefix=CHECK-MAX
 
 ; CHECK-MAX:        - Type:            MEMORY
 ; CHECK-MAX-NEXT:     Memories:
 ; CHECK-MAX-NEXT:       - Flags:           [ HAS_MAX ]

lld/wasm/SyntheticSections.h

 
 class MemorySection : public SyntheticSection {
 public:
   MemorySection() : SyntheticSection(llvm::wasm::WASM_SEC_MEMORY) {}
 
   bool isNeeded() const override { return !config->importMemory; }
   void writeBody() override;
 
-  uint32_t numMemoryPages = 0;
-  uint32_t maxMemoryPages = 0;
+  uint64_t numMemoryPages = 0;
+  uint64_t maxMemoryPages = 0;
 };
 
 // The event section contains a list of declared wasm events associated with the
 // module. Currently the only supported event kind is exceptions. A single event
 // entry represents a single event with an event tag. All C++ exceptions are
 // represented by a single event. An event entry in this section contains
 // information on what kind of event it is (e.g. exception) and the type of
 // values contained in a single event object. (In wasm, an event can contain

lld/wasm/SyntheticSections.cpp

if (config->importMemory) {
     import.Memory.Flags = 0;
     import.Memory.Initial = out.memorySec->numMemoryPages;
     if (out.memorySec->maxMemoryPages != 0 || config->sharedMemory) {
       import.Memory.Flags |= WASM_LIMITS_FLAG_HAS_MAX;
       import.Memory.Maximum = out.memorySec->maxMemoryPages;
     }
     if (config->sharedMemory)
       import.Memory.Flags |= WASM_LIMITS_FLAG_IS_SHARED;
+    if (config->is64)
+      import.Memory.Flags |= WASM_LIMITS_FLAG_IS_64;
     writeImport(os, import);
   }
 
   if (config->importTable) {
     uint32_t tableSize = config->tableBase + out.elemSec->numEntries();
     WasmImport import;
     import.Module = defaultModule;
     import.Field = functionTableName;
void MemorySection::writeBody() {
 
   bool hasMax = maxMemoryPages != 0 || config->sharedMemory;
   writeUleb128(os, 1, "memory count");
   unsigned flags = 0;
   if (hasMax)
     flags |= WASM_LIMITS_FLAG_HAS_MAX;
   if (config->sharedMemory)
     flags |= WASM_LIMITS_FLAG_IS_SHARED;
+  if (config->is64)
+    flags |= WASM_LIMITS_FLAG_IS_64;
   writeUleb128(os, flags, "memory limits flags");
   writeUleb128(os, numMemoryPages, "initial pages");
   if (hasMax)
     writeUleb128(os, maxMemoryPages, "max pages");
 }
 
 void EventSection::writeBody() {
   raw_ostream &os = bodyOutputStream;

lld/wasm/Writer.cpp

if (config->relocatable || config->isPic)
       return;
     memoryPtr = alignTo(memoryPtr, stackAlignment);
     if (config->zStackSize != alignTo(config->zStackSize, stackAlignment))
       error("stack size must be " + Twine(stackAlignment) + "-byte aligned");
     log("mem: stack size  = " + Twine(config->zStackSize));
     log("mem: stack base  = " + Twine(memoryPtr));
     memoryPtr += config->zStackSize;
     auto *sp = cast<DefinedGlobal>(WasmSym::stackPointer);
-    assert(sp->global->global.InitExpr.Opcode == WASM_OPCODE_I32_CONST);
+    switch (sp->global->global.InitExpr.Opcode) {
+    case WASM_OPCODE_I32_CONST:
       sp->global->global.InitExpr.Value.Int32 = memoryPtr;
+      break;
+    case WASM_OPCODE_I64_CONST:
+      sp->global->global.InitExpr.Value.Int64 = memoryPtr;
+      break;
+    default:
+      llvm_unreachable("init expr must be i32/i64.const");
+    }
     log("mem: stack top   = " + Twine(memoryPtr));
   };
 
   if (config->stackFirst) {
     placeStack();
   } else {
     memoryPtr = config->globalBase;
     log("mem: global base = " + Twine(config->globalBase));
void Writer::layoutMemory() {
 
   // Set `__heap_base` to directly follow the end of the stack or global data.
   // The fact that this comes last means that a malloc/brk implementation
   // can grow the heap at runtime.
   log("mem: heap base   = " + Twine(memoryPtr));
   if (WasmSym::heapBase)
     WasmSym::heapBase->setVirtualAddress(memoryPtr);
 
+  uint64_t maxMemorySetting = 1ULL << (config->is64 ? 48 : 32);
+
   if (config->initialMemory != 0) {
     if (config->initialMemory != alignTo(config->initialMemory, WasmPageSize))
       error("initial memory must be " + Twine(WasmPageSize) + "-byte aligned");
     if (memoryPtr > config->initialMemory)
       error("initial memory too small, " + Twine(memoryPtr) + " bytes needed");
-    if (config->initialMemory > (1ULL << 32))
-      error("initial memory too large, cannot be greater than 4294967296");
+    if (config->initialMemory > maxMemorySetting)
+      error("initial memory too large, cannot be greater than " +
+            Twine(maxMemorySetting));
     memoryPtr = config->initialMemory;
   }
   out.dylinkSec->memSize = memoryPtr;
   out.memorySec->numMemoryPages =
       alignTo(memoryPtr, WasmPageSize) / WasmPageSize;
   log("mem: total pages = " + Twine(out.memorySec->numMemoryPages));
 
   // Check max if explicitly supplied or required by shared memory
   if (config->maxMemory != 0 || config->sharedMemory) {
     if (config->maxMemory != alignTo(config->maxMemory, WasmPageSize))
       error("maximum memory must be " + Twine(WasmPageSize) + "-byte aligned");
     if (memoryPtr > config->maxMemory)
       error("maximum memory too small, " + Twine(memoryPtr) + " bytes needed");
-    if (config->maxMemory > (1ULL << 32))
-      error("maximum memory too large, cannot be greater than 4294967296");
+    if (config->maxMemory > maxMemorySetting)
+      error("maximum memory too large, cannot be greater than " +
+            Twine(maxMemorySetting));
     out.memorySec->maxMemoryPages = config->maxMemory / WasmPageSize;
     log("mem: max pages   = " + Twine(out.memorySec->maxMemoryPages));
   }
 }
 
 void Writer::addSection(OutputSection *sec) {
   if (!sec->isNeeded())
     return;

llvm/include/llvm/BinaryFormat/Wasm.h

 struct WasmExport {
   StringRef Name;
   uint8_t Kind;
   uint32_t Index;
 };
 
 struct WasmLimits {
   uint8_t Flags;
-  uint32_t Initial;
-  uint32_t Maximum;
+  uint64_t Initial;
+  uint64_t Maximum;
 };
 
 struct WasmTable {
   uint8_t ElemType;
   WasmLimits Limits;
 };
 
 struct WasmInitExpr {
enum : unsigned {
   WASM_OPCODE_I32_ATOMIC_WAIT = 0x01,
   WASM_OPCODE_I32_ATOMIC_STORE = 0x17,
   WASM_OPCODE_I32_RMW_CMPXCHG = 0x48,
 };
 
 enum : unsigned {
   WASM_LIMITS_FLAG_HAS_MAX = 0x1,
   WASM_LIMITS_FLAG_IS_SHARED = 0x2,
+  WASM_LIMITS_FLAG_IS_64 = 0x4,
 };
 
 enum : unsigned {
   WASM_SEGMENT_IS_PASSIVE = 0x01,
   WASM_SEGMENT_HAS_MEMINDEX = 0x02,
 };
 
 // Feature policy prefixes used in the custom "target_features" section

llvm/lib/MC/WasmObjectWriter.cpp

 // A wasm data segment.  A wasm binary contains only a single data section
 // but that can contain many segments, each with their own virtual location
 // in memory.  Each MCSection data created by llvm is modeled as its own
 // wasm data segment.
 struct WasmDataSegment {
   MCSectionWasm *Section;
   StringRef Name;
   uint32_t InitFlags;
-  uint32_t Offset;
+  uint64_t Offset;
   uint32_t Alignment;
   uint32_t LinkerFlags;
   SmallVector<char, 4> Data;
 };
 
 // A wasm function to be written into the function section.
 struct WasmFunction {
   uint32_t SigIndex;
void writeI64(int64_t val) {
     char Buffer[8];
     support::endian::write64le(Buffer, val);
     W.OS.write(Buffer, sizeof(Buffer));
   }
 
   void writeValueType(wasm::ValType Ty) { W.OS << static_cast<char>(Ty); }
 
   void writeTypeSection(ArrayRef<WasmSignature> Signatures);
-  void writeImportSection(ArrayRef<wasm::WasmImport> Imports, uint32_t DataSize,
+  void writeImportSection(ArrayRef<wasm::WasmImport> Imports, uint64_t DataSize,
                           uint32_t NumElements);
   void writeFunctionSection(ArrayRef<WasmFunction> Functions);
   void writeExportSection(ArrayRef<wasm::WasmExport> Exports);
   void writeElemSection(ArrayRef<uint32_t> TableElems);
   void writeDataCountSection();
   uint32_t writeCodeSection(const MCAssembler &Asm, const MCAsmLayout &Layout,
                             ArrayRef<WasmFunction> Functions);
   uint32_t writeDataSection(const MCAsmLayout &Layout);
for (const WasmSignature &Sig : Signatures) {
     for (wasm::ValType Ty : Sig.Returns)
       writeValueType(Ty);
   }
 
   endSection(Section);
 }
 
 void WasmObjectWriter::writeImportSection(ArrayRef<wasm::WasmImport> Imports,
-                                          uint32_t DataSize,
+                                          uint64_t DataSize,
                                           uint32_t NumElements) {
   if (Imports.empty())
     return;
 
-  uint32_t NumPages = (DataSize + wasm::WasmPageSize - 1) / wasm::WasmPageSize;
+  uint64_t NumPages = (DataSize + wasm::WasmPageSize - 1) / wasm::WasmPageSize;
 
   SectionBookkeeping Section;
   startSection(Section, wasm::WASM_SEC_IMPORT);
 
   encodeULEB128(Imports.size(), W.OS);
   for (const wasm::WasmImport &Import : Imports) {
     writeString(Import.Module);
     writeString(Import.Field);
     W.OS << char(Import.Kind);
 
     switch (Import.Kind) {
     case wasm::WASM_EXTERNAL_FUNCTION:
       encodeULEB128(Import.SigIndex, W.OS);
       break;
     case wasm::WASM_EXTERNAL_GLOBAL:
       W.OS << char(Import.Global.Type);
       W.OS << char(Import.Global.Mutable ? 1 : 0);
       break;
     case wasm::WASM_EXTERNAL_MEMORY:
-      encodeULEB128(0, W.OS);        // flags
-      encodeULEB128(NumPages, W.OS); // initial
+      encodeULEB128(Import.Memory.Flags, W.OS);
+      encodeULEB128(NumPages, W.OS);  // initial
       break;
     case wasm::WASM_EXTERNAL_TABLE:
       W.OS << char(Import.Table.ElemType);
       encodeULEB128(0, W.OS);           // flags
       encodeULEB128(NumElements, W.OS); // initial
       break;
     case wasm::WASM_EXTERNAL_EVENT:
       encodeULEB128(Import.Event.Attribute, W.OS);
uint32_t WasmObjectWriter::writeDataSection(const MCAsmLayout &Layout) {
 
   encodeULEB128(DataSegments.size(), W.OS); // count
 
   for (const WasmDataSegment &Segment : DataSegments) {
     encodeULEB128(Segment.InitFlags, W.OS); // flags
     if (Segment.InitFlags & wasm::WASM_SEGMENT_HAS_MEMINDEX)
       encodeULEB128(0, W.OS); // memory index
     if ((Segment.InitFlags & wasm::WASM_SEGMENT_IS_PASSIVE) == 0) {
-      W.OS << char(wasm::WASM_OPCODE_I32_CONST);
+      W.OS << char(Segment.Offset > std::numeric_limits<int32_t>().max()
+                     ? wasm::WASM_OPCODE_I64_CONST
+                     : wasm::WASM_OPCODE_I32_CONST);
       encodeSLEB128(Segment.Offset, W.OS); // offset
       W.OS << char(wasm::WASM_OPCODE_END);
     }
     encodeULEB128(Segment.Data.size(), W.OS); // size
     Segment.Section->setSectionOffset(W.OS.tell() - Section.ContentsOffset);
     W.OS << Segment.Data; // data
   }
 
uint64_t WasmObjectWriter::writeObject(MCAssembler &Asm,
   SmallVector<uint32_t, 4> TableElems;
   SmallVector<wasm::WasmImport, 4> Imports;
   SmallVector<wasm::WasmExport, 4> Exports;
   SmallVector<wasm::WasmEventType, 1> Events;
   SmallVector<wasm::WasmGlobal, 1> Globals;
   SmallVector<wasm::WasmSymbolInfo, 4> SymbolInfos;
   SmallVector<std::pair<uint16_t, uint32_t>, 2> InitFuncs;
   std::map<StringRef, std::vector<WasmComdatEntry>> Comdats;
-  uint32_t DataSize = 0;
+  uint64_t DataSize = 0;
 
   // For now, always emit the memory import, since loads and stores are not
   // valid without it. In the future, we could perhaps be more clever and omit
   // it if there are no loads or stores.
   wasm::WasmImport MemImport;
   MemImport.Module = "env";
   MemImport.Field = "__linear_memory";
   MemImport.Kind = wasm::WASM_EXTERNAL_MEMORY;
+  MemImport.Memory.Flags = is64Bit() ? wasm::WASM_LIMITS_FLAG_IS_64 : 0;
   Imports.push_back(MemImport);
 
   // For now, always emit the table section, since indirect calls are not
   // valid without it. In the future, we could perhaps be more clever and omit
   // it if there are no indirect calls.
   wasm::WasmImport TableImport;
   TableImport.Module = "env";
   TableImport.Field = "__indirect_function_table";

llvm/lib/Object/WasmObjectFile.cpp

return make_error<GenericBinaryError>("Invalid init_expr",
                                           object_error::parse_failed);
   }
   return Error::success();
 }
 
 static wasm::WasmLimits readLimits(WasmObjectFile::ReadContext &Ctx) {
   wasm::WasmLimits Result;
   Result.Flags = readVaruint32(Ctx);
-  Result.Initial = readVaruint32(Ctx);
+  Result.Initial = readVaruint64(Ctx);
   if (Result.Flags & wasm::WASM_LIMITS_FLAG_HAS_MAX)
-    Result.Maximum = readVaruint32(Ctx);
+    Result.Maximum = readVaruint64(Ctx);
   return Result;
 }
 
 static wasm::WasmTable readTable(WasmObjectFile::ReadContext &Ctx) {
   wasm::WasmTable Table;
   Table.ElemType = readUint8(Ctx);
   Table.Limits = readLimits(Ctx);
   return Table;

llvm/lib/ObjectYAML/WasmYAML.cpp

void MappingTraits<WasmYAML::Event>::mapping(IO &IO, WasmYAML::Event &Event) {
   IO.mapRequired("SigIndex", Event.SigIndex);
 }
 
 void ScalarBitSetTraits<WasmYAML::LimitFlags>::bitset(
     IO &IO, WasmYAML::LimitFlags &Value) {
 #define BCase(X) IO.bitSetCase(Value, #X, wasm::WASM_LIMITS_FLAG_##X)
   BCase(HAS_MAX);
   BCase(IS_SHARED);
+  BCase(IS_64);
 #undef BCase
 }
 
 void ScalarBitSetTraits<WasmYAML::SegmentFlags>::bitset(
     IO &IO, WasmYAML::SegmentFlags &Value) {}
 
 void ScalarBitSetTraits<WasmYAML::SymbolFlags>::bitset(
     IO &IO, WasmYAML::SymbolFlags &Value) {

llvm/test/MC/WebAssembly/wasm64.s

 # BIN-NEXT:           - I64
 # BIN-NEXT:         ReturnTypes:     []
 # BIN-NEXT:   - Type:            IMPORT
 # BIN-NEXT:     Imports:
 # BIN-NEXT:       - Module:          env
 # BIN-NEXT:         Field:           __linear_memory
 # BIN-NEXT:         Kind:            MEMORY
 # BIN-NEXT:         Memory:
+# BIN-NEXT:           Flags:           [ IS_64 ]
 # BIN-NEXT:           Initial:         0x00000001
 # BIN-NEXT:       - Module:          env
 # BIN-NEXT:         Field:           __indirect_function_table
 # BIN-NEXT:         Kind:            TABLE
 # BIN-NEXT:         Table:
 # BIN-NEXT:           ElemType:        FUNCREF
 # BIN-NEXT:           Limits:
 # BIN-NEXT:             Initial:         0x00000000