diff --git a/lld/COFF/Config.h b/lld/COFF/Config.h
--- a/lld/COFF/Config.h
+++ b/lld/COFF/Config.h
@@ -162,6 +162,8 @@
 
   // Used for /opt:lldlto=N
   unsigned ltoo = 2;
+  // Used for /opt:lldltocgo=N
+  std::optional<unsigned> ltocgo;
 
   // Used for /opt:lldltojobs=N
   std::string thinLTOJobs;
diff --git a/lld/COFF/Driver.cpp b/lld/COFF/Driver.cpp
--- a/lld/COFF/Driver.cpp
+++ b/lld/COFF/Driver.cpp
@@ -1771,6 +1771,11 @@
         StringRef optLevel = s.substr(7);
         if (optLevel.getAsInteger(10, config->ltoo) || config->ltoo > 3)
           error("/opt:lldlto: invalid optimization level: " + optLevel);
+      } else if (s.startswith("lldltocgo=")) {
+        StringRef optLevel = s.substr(10);
+        config->ltocgo.emplace();
+        if (optLevel.getAsInteger(10, *config->ltocgo) || *config->ltocgo > 3)
+          error("/opt:lldltocgo: invalid optimization level: " + optLevel);
       } else if (s.startswith("lldltojobs=")) {
         StringRef jobs = s.substr(11);
         if (!get_threadpool_strategy(jobs))
diff --git a/lld/COFF/LTO.cpp b/lld/COFF/LTO.cpp
--- a/lld/COFF/LTO.cpp
+++ b/lld/COFF/LTO.cpp
@@ -88,7 +88,8 @@
   c.OptLevel = ctx.config.ltoo;
   c.CPU = getCPUStr();
   c.MAttrs = getMAttrs();
-  c.CGOptLevel = args::getCGOptLevel(ctx.config.ltoo);
+  c.CGOptLevel =
+      *CodeGenOpt::getLevel(ctx.config.ltocgo.value_or(ctx.config.ltoo));
   c.AlwaysEmitRegularLTOObj = !ctx.config.ltoObjPath.empty();
   c.DebugPassManager = ctx.config.ltoDebugPassManager;
   c.CSIRProfile = std::string(ctx.config.ltoCSProfileFile);
diff --git a/lld/Common/Args.cpp b/lld/Common/Args.cpp
--- a/lld/Common/Args.cpp
+++ b/lld/Common/Args.cpp
@@ -17,15 +17,6 @@
 using namespace llvm;
 using namespace lld;
 
-// TODO(sbc): Remove this once CGOptLevel can be set completely based on bitcode
-// function metadata.
-CodeGenOpt::Level lld::args::getCGOptLevel(int optLevelLTO) {
-  if (optLevelLTO == 3)
-    return CodeGenOpt::Aggressive;
-  assert(optLevelLTO < 3);
-  return CodeGenOpt::Default;
-}
-
 static int64_t getInteger(opt::InputArgList &args, unsigned key,
                           int64_t Default, unsigned base) {
   auto *a = args.getLastArg(key);
diff --git a/lld/ELF/Config.h b/lld/ELF/Config.h
--- a/lld/ELF/Config.h
+++ b/lld/ELF/Config.h
@@ -318,6 +318,7 @@
   uint64_t zStackSize;
   unsigned ltoPartitions;
   unsigned ltoo;
+  llvm::CodeGenOpt::Level ltocgo;
   unsigned optimize;
   StringRef thinLTOJobs;
   unsigned timeTraceGranularity;
diff --git a/lld/ELF/Driver.cpp b/lld/ELF/Driver.cpp
--- a/lld/ELF/Driver.cpp
+++ b/lld/ELF/Driver.cpp
@@ -1139,6 +1139,8 @@
       args.hasFlag(OPT_lto_whole_program_visibility,
                    OPT_no_lto_whole_program_visibility, false);
   config->ltoo = args::getInteger(args, OPT_lto_O, 2);
+  config->ltocgo =
+      *CodeGenOpt::getLevel(args::getInteger(args, OPT_lto_CGO, config->ltoo));
   config->ltoObjPath = args.getLastArgValue(OPT_lto_obj_path_eq);
   config->ltoPartitions = args::getInteger(args, OPT_lto_partitions, 1);
   config->ltoSampleProfile = args.getLastArgValue(OPT_lto_sample_profile);
diff --git a/lld/ELF/LTO.cpp b/lld/ELF/LTO.cpp
--- a/lld/ELF/LTO.cpp
+++ b/lld/ELF/LTO.cpp
@@ -127,7 +127,7 @@
   c.OptLevel = config->ltoo;
   c.CPU = getCPUStr();
   c.MAttrs = getMAttrs();
-  c.CGOptLevel = args::getCGOptLevel(config->ltoo);
+  c.CGOptLevel = config->ltocgo;
 
   c.PTO.LoopVectorization = c.OptLevel > 1;
   c.PTO.SLPVectorization = c.OptLevel > 1;
diff --git a/lld/ELF/Options.td b/lld/ELF/Options.td
--- a/lld/ELF/Options.td
+++ b/lld/ELF/Options.td
@@ -560,6 +560,8 @@
   HelpText<"Passes to run during LTO">;
 def lto_O: JJ<"lto-O">, MetaVarName<"<opt-level>">,
   HelpText<"Optimization level for LTO">;
+def lto_CGO: JJ<"lto-CGO">, MetaVarName<"<cgopt-level>">,
+  HelpText<"Codegen optimization level for LTO">;
 def lto_partitions: JJ<"lto-partitions=">,
   HelpText<"Number of LTO codegen partitions">;
 def lto_cs_profile_generate: FF<"lto-cs-profile-generate">,
diff --git a/lld/MachO/Arch/ARM.cpp b/lld/MachO/Arch/ARM.cpp
--- a/lld/MachO/Arch/ARM.cpp
+++ b/lld/MachO/Arch/ARM.cpp
@@ -29,7 +29,7 @@
 
   int64_t getEmbeddedAddend(MemoryBufferRef, uint64_t offset,
                             const relocation_info) const override;
-  void relocateOne(uint8_t *loc, const Reloc &, uint64_t va,
+  void relocateOne(uint8_t *loc, const macho::Reloc &, uint64_t va,
                    uint64_t pc) const override;
 
   void writeStub(uint8_t *buf, const Symbol &, uint64_t) const override;
@@ -45,7 +45,7 @@
   void relaxGotLoad(uint8_t *loc, uint8_t type) const override;
   uint64_t getPageSize() const override { return 4 * 1024; }
 
-  void handleDtraceReloc(const Symbol *sym, const Reloc &r,
+  void handleDtraceReloc(const Symbol *sym, const macho::Reloc &r,
                          uint8_t *loc) const override;
 };
 } // namespace
@@ -101,7 +101,7 @@
 using Cond = Bitfield::Element<uint32_t, 28, 4>;
 using Imm24 = Bitfield::Element<int32_t, 0, 24>;
 
-void ARM::relocateOne(uint8_t *loc, const Reloc &r, uint64_t value,
+void ARM::relocateOne(uint8_t *loc, const macho::Reloc &r, uint64_t value,
                       uint64_t pc) const {
   switch (r.type) {
   case ARM_RELOC_BR24: {
@@ -180,7 +180,7 @@
   return &t;
 }
 
-void ARM::handleDtraceReloc(const Symbol *sym, const Reloc &r,
+void ARM::handleDtraceReloc(const Symbol *sym, const macho::Reloc &r,
                             uint8_t *loc) const {
   if (config->outputType == MH_OBJECT)
     return;
diff --git a/lld/MachO/Config.h b/lld/MachO/Config.h
--- a/lld/MachO/Config.h
+++ b/lld/MachO/Config.h
@@ -18,6 +18,7 @@
 #include "llvm/ADT/StringSet.h"
 #include "llvm/BinaryFormat/MachO.h"
 #include "llvm/Support/CachePruning.h"
+#include "llvm/Support/CodeGen.h"
 #include "llvm/Support/GlobPattern.h"
 #include "llvm/Support/VersionTuple.h"
 #include "llvm/TextAPI/Architecture.h"
@@ -165,6 +166,7 @@
   llvm::StringRef thinLTOJobs;
   llvm::StringRef umbrella;
   uint32_t ltoo = 2;
+  llvm::CodeGenOpt::Level ltocgo;
   llvm::CachePruningPolicy thinLTOCachePolicy;
   llvm::StringRef thinLTOCacheDir;
   llvm::StringRef thinLTOIndexOnlyArg;
diff --git a/lld/MachO/Driver.cpp b/lld/MachO/Driver.cpp
--- a/lld/MachO/Driver.cpp
+++ b/lld/MachO/Driver.cpp
@@ -1561,6 +1561,8 @@
   config->ltoo = args::getInteger(args, OPT_lto_O, 2);
   if (config->ltoo > 3)
     error("--lto-O: invalid optimization level: " + Twine(config->ltoo));
+  config->ltocgo =
+      *CodeGenOpt::getLevel(args::getInteger(args, OPT_lto_CGO, config->ltoo));
   config->thinLTOCacheDir = args.getLastArgValue(OPT_cache_path_lto);
   config->thinLTOCachePolicy = getLTOCachePolicy(args);
   config->thinLTOEmitImportsFiles = args.hasArg(OPT_thinlto_emit_imports_files);
diff --git a/lld/MachO/EhFrame.cpp b/lld/MachO/EhFrame.cpp
--- a/lld/MachO/EhFrame.cpp
+++ b/lld/MachO/EhFrame.cpp
@@ -109,16 +109,16 @@
 static void createSubtraction(PointerUnion<Symbol *, InputSection *> a,
                               PointerUnion<Symbol *, InputSection *> b,
                               uint64_t off, uint8_t length,
-                              SmallVectorImpl<Reloc> *newRelocs) {
+                              SmallVectorImpl<macho::Reloc> *newRelocs) {
   auto subtrahend = a;
   auto minuend = b;
   if (Invert)
     std::swap(subtrahend, minuend);
   assert(subtrahend.is<Symbol *>());
-  Reloc subtrahendReloc(target->subtractorRelocType, /*pcrel=*/false, length,
-                        off, /*addend=*/0, subtrahend);
-  Reloc minuendReloc(target->unsignedRelocType, /*pcrel=*/false, length, off,
-                     (Invert ? 1 : -1) * off, minuend);
+  macho::Reloc subtrahendReloc(target->subtractorRelocType, /*pcrel=*/false,
+                               length, off, /*addend=*/0, subtrahend);
+  macho::Reloc minuendReloc(target->unsignedRelocType, /*pcrel=*/false, length,
+                            off, (Invert ? 1 : -1) * off, minuend);
   newRelocs->push_back(subtrahendReloc);
   newRelocs->push_back(minuendReloc);
 }
diff --git a/lld/MachO/ICF.cpp b/lld/MachO/ICF.cpp
--- a/lld/MachO/ICF.cpp
+++ b/lld/MachO/ICF.cpp
@@ -105,7 +105,7 @@
     return false;
   if (ia->relocs.size() != ib->relocs.size())
     return false;
-  auto f = [](const Reloc &ra, const Reloc &rb) {
+  auto f = [](const macho::Reloc &ra, const macho::Reloc &rb) {
     if (ra.type != rb.type)
       return false;
     if (ra.pcrel != rb.pcrel)
@@ -178,7 +178,7 @@
   if (verboseDiagnostics)
     ++equalsVariableCount;
   assert(ia->relocs.size() == ib->relocs.size());
-  auto f = [this](const Reloc &ra, const Reloc &rb) {
+  auto f = [this](const macho::Reloc &ra, const macho::Reloc &rb) {
     // We already filtered out mismatching values/addends in equalsConstant.
     if (ra.referent == rb.referent)
       return true;
@@ -286,7 +286,7 @@
   for (icfPass = 0; icfPass < 2; ++icfPass) {
     parallelForEach(icfInputs, [&](ConcatInputSection *isec) {
       uint32_t hash = isec->icfEqClass[icfPass % 2];
-      for (const Reloc &r : isec->relocs) {
+      for (const macho::Reloc &r : isec->relocs) {
         if (auto *sym = r.referent.dyn_cast<Symbol *>()) {
           if (auto *defined = dyn_cast<Defined>(sym)) {
             if (defined->isec) {
@@ -386,7 +386,7 @@
 
     const InputSection *isec = addrSigSection->subsections[0].isec;
 
-    for (const Reloc &r : isec->relocs) {
+    for (const macho::Reloc &r : isec->relocs) {
       if (auto *sym = r.referent.dyn_cast<Symbol *>())
         markSymAsAddrSig(sym);
       else
@@ -442,7 +442,7 @@
         // We have to do this copying serially as the BumpPtrAllocator is not
         // thread-safe. FIXME: Make a thread-safe allocator.
         MutableArrayRef<uint8_t> copy = isec->data.copy(bAlloc());
-        for (const Reloc &r : isec->relocs)
+        for (const macho::Reloc &r : isec->relocs)
           target->relocateOne(copy.data() + r.offset, r, /*va=*/0,
                               /*relocVA=*/0);
         isec->data = copy;
diff --git a/lld/MachO/InputSection.cpp b/lld/MachO/InputSection.cpp
--- a/lld/MachO/InputSection.cpp
+++ b/lld/MachO/InputSection.cpp
@@ -31,8 +31,8 @@
 // Verify ConcatInputSection's size on 64-bit builds. The size of std::vector
 // can differ based on STL debug levels (e.g. iterator debugging on MSVC's STL),
 // so account for that.
-static_assert(sizeof(void *) != 8 ||
-                  sizeof(ConcatInputSection) == sizeof(std::vector<Reloc>) + 88,
+static_assert(sizeof(void *) != 8 || sizeof(ConcatInputSection) ==
+                                         sizeof(std::vector<macho::Reloc>) + 88,
               "Try to minimize ConcatInputSection's size, we create many "
               "instances of it");
 
@@ -166,7 +166,7 @@
   memcpy(buf, data.data(), data.size());
 
   for (size_t i = 0; i < relocs.size(); i++) {
-    const Reloc &r = relocs[i];
+    const macho::Reloc &r = relocs[i];
     uint8_t *loc = buf + r.offset;
     uint64_t referentVA = 0;
 
@@ -174,7 +174,7 @@
                             target->hasAttr(r.type, RelocAttrBits::UNSIGNED);
     if (target->hasAttr(r.type, RelocAttrBits::SUBTRAHEND)) {
       const Symbol *fromSym = r.referent.get<Symbol *>();
-      const Reloc &minuend = relocs[++i];
+      const macho::Reloc &minuend = relocs[++i];
       uint64_t minuendVA;
       if (const Symbol *toSym = minuend.referent.dyn_cast<Symbol *>())
         minuendVA = toSym->getVA() + minuend.addend;
diff --git a/lld/MachO/LTO.cpp b/lld/MachO/LTO.cpp
--- a/lld/MachO/LTO.cpp
+++ b/lld/MachO/LTO.cpp
@@ -70,7 +70,7 @@
   c.TimeTraceEnabled = config->timeTraceEnabled;
   c.TimeTraceGranularity = config->timeTraceGranularity;
   c.OptLevel = config->ltoo;
-  c.CGOptLevel = args::getCGOptLevel(config->ltoo);
+  c.CGOptLevel = config->ltocgo;
   if (config->saveTemps)
     checkError(c.addSaveTemps(config->outputFile.str() + ".",
                               /*UseInputModulePath=*/true));
diff --git a/lld/MachO/Options.td b/lld/MachO/Options.td
--- a/lld/MachO/Options.td
+++ b/lld/MachO/Options.td
@@ -85,6 +85,10 @@
     Group<grp_lld>;
 def lto_O: Joined<["--"], "lto-O">,
     HelpText<"Set optimization level for LTO (default: 2)">,
+    MetaVarName<"<cgopt-level>">,
+    Group<grp_lld>;
+def lto_CGO: Joined<["--"], "lto-CGO">,
+    HelpText<"Set codegen optimization level for LTO (default: 2)">,
     MetaVarName<"<opt-level>">,
     Group<grp_lld>;
 def thinlto_cache_policy_eq: Joined<["--"], "thinlto-cache-policy=">,
diff --git a/lld/MachO/Relocations.cpp b/lld/MachO/Relocations.cpp
--- a/lld/MachO/Relocations.cpp
+++ b/lld/MachO/Relocations.cpp
@@ -18,11 +18,12 @@
 using namespace lld;
 using namespace lld::macho;
 
-static_assert(sizeof(void *) != 8 || sizeof(Reloc) == 24,
+static_assert(sizeof(void *) != 8 || sizeof(macho::Reloc) == 24,
               "Try to minimize Reloc's size; we create many instances");
 
 bool macho::validateSymbolRelocation(const Symbol *sym,
-                                     const InputSection *isec, const Reloc &r) {
+                                     const InputSection *isec,
+                                     const macho::Reloc &r) {
   const RelocAttrs &relocAttrs = target->getRelocAttrs(r.type);
   bool valid = true;
   auto message = [&](const Twine &diagnostic) {
@@ -82,7 +83,7 @@
   return nullptr;
 }
 
-void macho::reportRangeError(void *loc, const Reloc &r, const Twine &v,
+void macho::reportRangeError(void *loc, const macho::Reloc &r, const Twine &v,
                              uint8_t bits, int64_t min, uint64_t max) {
   std::string hint;
   uint64_t off = reinterpret_cast<const uint8_t *>(loc) - in.bufferStart;
diff --git a/lld/MachO/UnwindInfoSection.cpp b/lld/MachO/UnwindInfoSection.cpp
--- a/lld/MachO/UnwindInfoSection.cpp
+++ b/lld/MachO/UnwindInfoSection.cpp
@@ -249,7 +249,7 @@
   // that are referenced from many places, at least some of them likely
   // live, it wouldn't reduce number of got entries.
   for (size_t i = 0; i < isec->relocs.size(); ++i) {
-    Reloc &r = isec->relocs[i];
+    macho::Reloc &r = isec->relocs[i];
     assert(target->hasAttr(r.type, RelocAttrBits::UNSIGNED));
     // Since compact unwind sections aren't part of the inputSections vector,
     // they don't get canonicalized by scanRelocations(), so we have to do the
@@ -375,7 +375,7 @@
     cu.functionLength =
         support::endian::read32le(buf + cuOffsets.functionLength);
     cu.encoding = support::endian::read32le(buf + cuOffsets.encoding);
-    for (const Reloc &r : d->unwindEntry->relocs) {
+    for (const macho::Reloc &r : d->unwindEntry->relocs) {
       if (r.offset == cuOffsets.personality) {
         cu.personality = r.referent.get<Symbol *>();
       } else if (r.offset == cuOffsets.lsda) {
diff --git a/lld/include/lld/Common/Args.h b/lld/include/lld/Common/Args.h
--- a/lld/include/lld/Common/Args.h
+++ b/lld/include/lld/Common/Args.h
@@ -23,8 +23,6 @@
 namespace lld {
 namespace args {
 
-llvm::CodeGenOpt::Level getCGOptLevel(int optLevelLTO);
-
 int64_t getInteger(llvm::opt::InputArgList &args, unsigned key,
                    int64_t Default);
 
diff --git a/lld/test/COFF/lto-cgo.ll b/lld/test/COFF/lto-cgo.ll
new file mode 100644
--- /dev/null
+++ b/lld/test/COFF/lto-cgo.ll
@@ -0,0 +1,22 @@
+; REQUIRES: x86
+
+; RUN: llvm-as %s -o %t.obj
+; RUN: lld-link -opt:lldlto=0 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: lld-link -opt:lldltocgo=0 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: lld-link -opt:lldlto=3 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: lld-link -opt:lldltocgo=3 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: lld-link -opt:lldlto=0 -opt:lldltocgo=0 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: lld-link -opt:lldlto=3 -opt:lldltocgo=0 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: lld-link -opt:lldlto=0 -opt:lldltocgo=3 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: lld-link -opt:lldlto=3 -opt:lldltocgo=3 %t.obj -dll -noentry -out:%t.dll -mllvm:-debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+
+; NOOPT: Fast Register Allocator
+; OPT: Greedy Register Allocator
+
+target datalayout = "e-m:w-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-pc-windows-msvc"
+
+define void @_start() {
+entry:
+  ret void
+}
diff --git a/lld/test/ELF/lto/cgo.ll b/lld/test/ELF/lto/cgo.ll
new file mode 100644
--- /dev/null
+++ b/lld/test/ELF/lto/cgo.ll
@@ -0,0 +1,22 @@
+; REQUIRES: x86
+
+; RUN: llvm-as %s -o %t.o
+; RUN: ld.lld --lto-O0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: ld.lld --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: ld.lld --lto-O3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: ld.lld --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: ld.lld --lto-O0 --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: ld.lld --lto-O3 --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: ld.lld --lto-O0 --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: ld.lld --lto-O3 --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+
+; NOOPT: Fast Register Allocator
+; OPT: Greedy Register Allocator
+
+target triple = "x86_64-unknown-linux-gnu"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+
+define void @_start() {
+entry:
+  ret void
+}
diff --git a/lld/test/MachO/lto-cgo.ll b/lld/test/MachO/lto-cgo.ll
new file mode 100644
--- /dev/null
+++ b/lld/test/MachO/lto-cgo.ll
@@ -0,0 +1,22 @@
+; REQUIRES: x86
+
+; RUN: llvm-as %s -o %t.o
+; RUN: %lld -dylib --lto-O0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: %lld -dylib --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: %lld -dylib --lto-O3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: %lld -dylib --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: %lld -dylib --lto-O0 --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: %lld -dylib --lto-O3 --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: %lld -dylib --lto-O0 --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: %lld -dylib --lto-O3 --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+
+; NOOPT: Fast Register Allocator
+; OPT: Greedy Register Allocator
+
+target triple = "x86_64-apple-darwin"
+target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+
+define void @_start() {
+entry:
+  ret void
+}
diff --git a/lld/test/wasm/lto/cgo.ll b/lld/test/wasm/lto/cgo.ll
new file mode 100644
--- /dev/null
+++ b/lld/test/wasm/lto/cgo.ll
@@ -0,0 +1,22 @@
+; REQUIRES: x86
+
+; RUN: llvm-as %s -o %t.o
+; RUN: wasm-ld --lto-O0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: wasm-ld --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: wasm-ld --lto-O3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: wasm-ld --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: wasm-ld --lto-O0 --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: wasm-ld --lto-O3 --lto-CGO0 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=NOOPT %s
+; RUN: wasm-ld --lto-O0 --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+; RUN: wasm-ld --lto-O3 --lto-CGO3 %t.o -o %t -mllvm -debug-pass=Structure 2>&1 | FileCheck --check-prefix=OPT %s
+
+; NOOPT-NOT: WebAssembly Optimize Returned
+; OPT: WebAssembly Optimize Returned
+
+target datalayout = "e-m:e-p:32:32-p10:8:8-p20:8:8-i64:64-n32:64-S128"
+target triple = "wasm32-unknown-unknown-wasm"
+
+define void @_start() {
+entry:
+  ret void
+}
diff --git a/lld/wasm/Config.h b/lld/wasm/Config.h
--- a/lld/wasm/Config.h
+++ b/lld/wasm/Config.h
@@ -14,6 +14,7 @@
 #include "llvm/ADT/StringSet.h"
 #include "llvm/BinaryFormat/Wasm.h"
 #include "llvm/Support/CachePruning.h"
+#include "llvm/Support/CodeGen.h"
 #include <optional>
 
 namespace lld {
@@ -63,6 +64,7 @@
   uint64_t zStackSize;
   unsigned ltoPartitions;
   unsigned ltoo;
+  llvm::CodeGenOpt::Level ltocgo;
   unsigned optimize;
   llvm::StringRef thinLTOJobs;
   bool ltoDebugPassManager;
diff --git a/lld/wasm/Driver.cpp b/lld/wasm/Driver.cpp
--- a/lld/wasm/Driver.cpp
+++ b/lld/wasm/Driver.cpp
@@ -424,6 +424,8 @@
   config->importTable = args.hasArg(OPT_import_table);
   config->importUndefined = args.hasArg(OPT_import_undefined);
   config->ltoo = args::getInteger(args, OPT_lto_O, 2);
+  config->ltocgo =
+      *CodeGenOpt::getLevel(args::getInteger(args, OPT_lto_CGO, config->ltoo));
   config->ltoPartitions = args::getInteger(args, OPT_lto_partitions, 1);
   config->ltoDebugPassManager = args.hasArg(OPT_lto_debug_pass_manager);
   config->mapFile = args.getLastArgValue(OPT_Map);
diff --git a/lld/wasm/LTO.cpp b/lld/wasm/LTO.cpp
--- a/lld/wasm/LTO.cpp
+++ b/lld/wasm/LTO.cpp
@@ -51,7 +51,7 @@
   c.DiagHandler = diagnosticHandler;
   c.OptLevel = config->ltoo;
   c.MAttrs = getMAttrs();
-  c.CGOptLevel = args::getCGOptLevel(config->ltoo);
+  c.CGOptLevel = config->ltocgo;
   c.DebugPassManager = config->ltoDebugPassManager;
 
   if (config->relocatable)
diff --git a/lld/wasm/Options.td b/lld/wasm/Options.td
--- a/lld/wasm/Options.td
+++ b/lld/wasm/Options.td
@@ -256,6 +256,8 @@
 // LTO-related options.
 def lto_O: JJ<"lto-O">, MetaVarName<"<opt-level>">,
   HelpText<"Optimization level for LTO">;
+def lto_CGO: JJ<"lto-CGO">, MetaVarName<"<cgopt-level>">,
+  HelpText<"Codegen optimization level for LTO">;
 def lto_partitions: JJ<"lto-partitions=">,
   HelpText<"Number of LTO codegen partitions">;
 def disable_verify: F<"disable-verify">;