Index: test/wasm/compress-relocs.ll
===================================================================
--- /dev/null
+++ test/wasm/compress-relocs.ll
@@ -0,0 +1,22 @@
+; RUN: llc -filetype=obj %p/Inputs/call-indirect.ll -o %t2.o
+; RUN: llc -filetype=obj %s -o %t.o
+; RUN: wasm-ld --check-signatures -o %t.wasm %t2.o %t.o
+; RUN: obj2yaml %t.wasm | FileCheck %s
+
+; RUN: wasm-ld --check-signatures --compress-relocs -o %t-compressed.wasm %t2.o %t.o
+; RUN: obj2yaml %t-compressed.wasm | FileCheck %s -check-prefix=COMPRESS
+
+target triple = "wasm32-unknown-unknown-wasm"
+
+define i32 @foo() {
+entry:
+  ret i32 2
+}
+
+define void @_start() local_unnamed_addr {
+entry:
+  ret void
+}
+
+; CHECK:    Body:            4100280284888080002100410028028088808000118080808000001A2000118180808000001A0B
+; COMPRESS: Body:            41002802840821004100280280081100001A20001101001A0B
Index: wasm/Config.h
===================================================================
--- wasm/Config.h
+++ wasm/Config.h
@@ -20,6 +20,7 @@
 struct Configuration {
   bool AllowUndefined;
   bool CheckSignatures;
+  bool CompressRelocTargets;
   bool Demangle;
   bool ExportTable;
   bool GcSections;
Index: wasm/Driver.cpp
===================================================================
--- wasm/Driver.cpp
+++ wasm/Driver.cpp
@@ -282,6 +282,8 @@
   errorHandler().ErrorLimit = args::getInteger(Args, OPT_error_limit, 20);
 
   Config->AllowUndefined = Args.hasArg(OPT_allow_undefined);
+  Config->CompressRelocTargets =
+      Args.hasFlag(OPT_compress_relocs, OPT_no_compress_relocs, false);
   Config->CheckSignatures =
       Args.hasFlag(OPT_check_signatures, OPT_no_check_signatures, false);
   Config->Demangle = Args.hasFlag(OPT_demangle, OPT_no_demangle, true);
@@ -308,6 +310,10 @@
   Config->ZStackSize =
       args::getZOptionValue(Args, OPT_z, "stack-size", WasmPageSize);
 
+  // We can't compress relocation targets when writing a relocatable file.
+  if (Config->Relocatable)
+    Config->CompressRelocTargets = false;
+
   if (auto *Arg = Args.getLastArg(OPT_allow_undefined_file))
     readImportFile(Arg->getValue());
 
Index: wasm/InputChunks.h
===================================================================
--- wasm/InputChunks.h
+++ wasm/InputChunks.h
@@ -48,11 +48,11 @@
 
   Kind kind() const { return SectionKind; }
 
-  uint32_t getSize() const { return data().size(); }
+  virtual uint32_t getSize() const { return data().size(); }
 
   void copyRelocations(const WasmSection &Section);
 
-  void writeTo(uint8_t *SectionStart) const;
+  virtual void writeTo(uint8_t *SectionStart) const;
 
   ArrayRef<WasmRelocation> getRelocations() const { return Relocations; }
 
@@ -78,6 +78,7 @@
   virtual ~InputChunk() = default;
   virtual ArrayRef<uint8_t> data() const = 0;
   virtual uint32_t getInputSectionOffset() const = 0;
+  virtual uint32_t getInputSize() const { return getSize(); };
 
   std::vector<WasmRelocation> Relocations;
   Kind SectionKind;
@@ -127,23 +128,34 @@
            C->kind() == InputChunk::SyntheticFunction;
   }
 
+  void writeTo(uint8_t *SectionStart) const override;
   StringRef getName() const override { return Function->SymbolName; }
   StringRef getDebugName() const override { return Function->DebugName; }
   uint32_t getComdat() const override { return Function->Comdat; }
+  uint32_t getSize() const override {
+    return Config->CompressRelocTargets && File ? CompressedSize : data().size();
+  }
   uint32_t getFunctionIndex() const { return FunctionIndex.getValue(); }
   bool hasFunctionIndex() const { return FunctionIndex.hasValue(); }
   void setFunctionIndex(uint32_t Index);
   uint32_t getTableIndex() const { return TableIndex.getValue(); }
   bool hasTableIndex() const { return TableIndex.hasValue(); }
   void setTableIndex(uint32_t Index);
+  void calculateSize();
 
   const WasmSignature &Signature;
 
 protected:
+  void compressRelocations();
+
   ArrayRef<uint8_t> data() const override {
+    assert(!Config->CompressRelocTargets);
     return File->CodeSection->Content.slice(getInputSectionOffset(),
                                             Function->Size);
   }
+
+  uint32_t getInputSize() const override { return Function->Size; }
+
   uint32_t getInputSectionOffset() const override {
     return Function->CodeSectionOffset;
   }
@@ -151,6 +163,10 @@
   const WasmFunction *Function;
   llvm::Optional<uint32_t> FunctionIndex;
   llvm::Optional<uint32_t> TableIndex;
+
+  std::vector<uint32_t> RelocationDeltas;
+  unsigned CompressedFuncSize = 0;
+  unsigned CompressedSize = 0;
 };
 
 class SyntheticFunction : public InputFunction {
Index: wasm/InputChunks.cpp
===================================================================
--- wasm/InputChunks.cpp
+++ wasm/InputChunks.cpp
@@ -47,7 +47,7 @@
   if (Section.Relocations.empty())
     return;
   size_t Start = getInputSectionOffset();
-  size_t Size = getSize();
+  size_t Size = getInputSize();
   for (const WasmRelocation &R : Section.Relocations)
     if (R.Offset >= Start && R.Offset < Start + Size)
       Relocations.push_back(R);
@@ -143,3 +143,112 @@
   assert(!hasTableIndex());
   TableIndex = Index;
 }
+
+static unsigned getRelocWidthPadded(const WasmRelocation &Rel) {
+  switch (Rel.Type) {
+  case R_WEBASSEMBLY_TYPE_INDEX_LEB:
+  case R_WEBASSEMBLY_FUNCTION_INDEX_LEB:
+  case R_WEBASSEMBLY_GLOBAL_INDEX_LEB:
+  case R_WEBASSEMBLY_MEMORY_ADDR_LEB:
+  case R_WEBASSEMBLY_TABLE_INDEX_SLEB:
+  case R_WEBASSEMBLY_MEMORY_ADDR_SLEB:
+    return 5;
+  case R_WEBASSEMBLY_TABLE_INDEX_I32:
+  case R_WEBASSEMBLY_MEMORY_ADDR_I32:
+    return 4;
+  default:
+    llvm_unreachable("unknown relocation type");
+  }
+}
+
+static unsigned writeCompressedReloc(const WasmRelocation &Rel, uint8_t *Buf,
+                                     uint32_t Value) {
+  switch (Rel.Type) {
+  case R_WEBASSEMBLY_TYPE_INDEX_LEB:
+  case R_WEBASSEMBLY_FUNCTION_INDEX_LEB:
+  case R_WEBASSEMBLY_GLOBAL_INDEX_LEB:
+  case R_WEBASSEMBLY_MEMORY_ADDR_LEB:
+    return encodeULEB128(Value, Buf);
+  case R_WEBASSEMBLY_TABLE_INDEX_SLEB:
+  case R_WEBASSEMBLY_MEMORY_ADDR_SLEB:
+    return encodeSLEB128(static_cast<int32_t>(Value), Buf);
+  case R_WEBASSEMBLY_TABLE_INDEX_I32:
+  case R_WEBASSEMBLY_MEMORY_ADDR_I32:
+    return 4;
+  default:
+    llvm_unreachable("unknown relocation type");
+  }
+}
+
+static unsigned getRelocWidth(const WasmRelocation &Rel, uint32_t Value) {
+  uint8_t Target[5];
+  return writeCompressedReloc(Rel, Target, Value);
+}
+
+void InputFunction::calculateSize() {
+  if (!File || !Config->CompressRelocTargets)
+    return;
+
+  DEBUG(dbgs() << "calculateSize: " << getName() << "\n");
+
+  const uint8_t *SecStart = File->CodeSection->Content.data();
+  const uint8_t *FuncStart = SecStart + getInputSectionOffset();
+  uint32_t FunctionSizeLength;
+  decodeULEB128(FuncStart, &FunctionSizeLength);
+
+  uint32_t Start = getInputSectionOffset();
+  uint32_t End = Start + Function->Size;
+
+  uint32_t LastRelocEnd = Start + FunctionSizeLength;
+  for (WasmRelocation &Rel : Relocations) {
+    DEBUG(dbgs() << "  region: " << (Rel.Offset - LastRelocEnd) << "\n");
+    assert(Rel.Offset >= LastRelocEnd);
+    CompressedFuncSize += Rel.Offset - LastRelocEnd;
+    CompressedFuncSize += getRelocWidth(Rel, File->calcNewValue(Rel));
+    LastRelocEnd = Rel.Offset + getRelocWidthPadded(Rel);
+    assert(LastRelocEnd <= End);
+  }
+  DEBUG(dbgs() << "  final region: " << (End - LastRelocEnd) << "\n");
+  CompressedFuncSize += End - LastRelocEnd;
+
+  // Now we know how long the resulting function is we can add the encoding
+  // of its length
+  uint8_t Buf[5];
+  CompressedSize = CompressedFuncSize + encodeULEB128(CompressedFuncSize, Buf);
+
+  DEBUG(dbgs() << "  calculateSize orig: " << Function->Size << "\n");
+  DEBUG(dbgs() << "  calculateSize  new: " << CompressedSize << "\n");
+}
+
+void InputFunction::writeTo(uint8_t *Buf) const {
+  if (!File || !Config->CompressRelocTargets)
+    return InputChunk::writeTo(Buf);
+
+  Buf += OutputOffset;
+  uint8_t *Orig = Buf;
+
+  const uint8_t *SecStart = File->CodeSection->Content.data();
+  const uint8_t *FuncStart = SecStart + getInputSectionOffset();
+  const uint8_t *End = FuncStart + Function->Size;
+  uint32_t Count;
+  decodeULEB128(Buf, &Count);
+  FuncStart += Count;
+
+  DEBUG(dbgs() << "write func: " << getName() << "\n");
+  Buf += encodeULEB128(CompressedFuncSize, Buf);
+  const uint8_t *LastRelocEnd = FuncStart;
+  for (const WasmRelocation &Rel : Relocations) {
+    unsigned ChunkSize = (SecStart + Rel.Offset) - LastRelocEnd;
+    DEBUG(dbgs() << "  write chunk: " << ChunkSize << "\n");
+    memcpy(Buf, LastRelocEnd, ChunkSize);
+    Buf += ChunkSize;
+    Buf += writeCompressedReloc(Rel, Buf, File->calcNewValue(Rel));
+    LastRelocEnd = SecStart + Rel.Offset + getRelocWidthPadded(Rel);
+  }
+
+  assert(LastRelocEnd <= End);
+  unsigned ChunkSize = End - LastRelocEnd;
+  DEBUG(dbgs() << "  write final chunk: " << ChunkSize << "\n");
+  memcpy(Buf, LastRelocEnd, ChunkSize);
+  DEBUG(dbgs() << "  total: " << (Buf + ChunkSize - Orig) << "\n");
+}
Index: wasm/Options.td
===================================================================
--- wasm/Options.td
+++ wasm/Options.td
@@ -23,6 +23,10 @@
 def color_diagnostics_eq: J<"color-diagnostics=">,
   HelpText<"Use colors in diagnostics">;
 
+defm compress_relocs: B<"compress-relocs",
+    "Compress relocation targets that contain padding",
+    "Do not compress relocation targets that contain padding">;
+
 defm demangle: B<"demangle",
     "Demangle symbol names",
     "Do not demangle symbol names">;
Index: wasm/OutputSections.cpp
===================================================================
--- wasm/OutputSections.cpp
+++ wasm/OutputSections.cpp
@@ -85,8 +85,9 @@
   OS.flush();
   BodySize = CodeSectionHeader.size();
 
-  for (InputChunk *Func : Functions) {
+  for (InputFunction *Func : Functions) {
     Func->OutputOffset = BodySize;
+    Func->calculateSize();
     BodySize += Func->getSize();
   }