Index: llvm/lib/Target/AMDGPU/AMDGPU.h
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPU.h
+++ llvm/lib/Target/AMDGPU/AMDGPU.h
@@ -16,6 +16,7 @@
 
 namespace llvm {
 
+class AMDGPUTargetMachine;
 class TargetMachine;
 
 // GlobalISel passes
@@ -53,7 +54,8 @@
 FunctionPass *createAMDGPULateCodeGenPreparePass();
 FunctionPass *createAMDGPUMachineCFGStructurizerPass();
 FunctionPass *createAMDGPURewriteOutArgumentsPass();
-ModulePass *createAMDGPULowerModuleLDSPass();
+ModulePass *
+createAMDGPULowerModuleLDSLegacyPass(const AMDGPUTargetMachine *TM = nullptr);
 FunctionPass *createSIModeRegisterPass();
 FunctionPass *createGCNPreRAOptimizationsPass();
 
@@ -112,10 +114,13 @@
   PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
 };
 
-void initializeAMDGPULowerModuleLDSPass(PassRegistry &);
-extern char &AMDGPULowerModuleLDSID;
+void initializeAMDGPULowerModuleLDSLegacyPass(PassRegistry &);
+extern char &AMDGPULowerModuleLDSLegacyPassID;
 
 struct AMDGPULowerModuleLDSPass : PassInfoMixin<AMDGPULowerModuleLDSPass> {
+  const AMDGPUTargetMachine &TM;
+  AMDGPULowerModuleLDSPass(const AMDGPUTargetMachine &TM_) : TM(TM_) {}
+
   PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
 };
 
Index: llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
@@ -177,6 +177,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPU.h"
+#include "AMDGPUTargetMachine.h"
 #include "Utils/AMDGPUBaseInfo.h"
 #include "Utils/AMDGPUMemoryUtils.h"
 #include "llvm/ADT/BitVector.h"
@@ -186,6 +187,7 @@
 #include "llvm/ADT/SetOperations.h"
 #include "llvm/ADT/SetVector.h"
 #include "llvm/Analysis/CallGraph.h"
+#include "llvm/CodeGen/TargetPassConfig.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DerivedTypes.h"
 #include "llvm/IR/IRBuilder.h"
@@ -252,7 +254,8 @@
   return {std::move(V)};
 }
 
-class AMDGPULowerModuleLDS : public ModulePass {
+class AMDGPULowerModuleLDS {
+  const AMDGPUTargetMachine &TM;
 
   static void
   removeLocalVarsFromUsedLists(Module &M,
@@ -326,11 +329,7 @@
   }
 
 public:
-  static char ID;
-
-  AMDGPULowerModuleLDS() : ModulePass(ID) {
-    initializeAMDGPULowerModuleLDSPass(*PassRegistry::getPassRegistry());
-  }
+  AMDGPULowerModuleLDS(const AMDGPUTargetMachine &TM_) : TM(TM_) {}
 
   using FunctionVariableMap = DenseMap<Function *, DenseSet<GlobalVariable *>>;
 
@@ -1089,7 +1088,7 @@
     return KernelToCreatedDynamicLDS;
   }
 
-  bool runOnModule(Module &M) override {
+  bool runOnModule(Module &M) {
     CallGraph CG = CallGraph(M);
     bool Changed = superAlignLDSGlobals(M);
 
@@ -1241,6 +1240,7 @@
         }
 
         if (Offset != 0) {
+          (void)TM; // TODO: Account for target maximum LDS
           std::string Buffer;
           raw_string_ostream SS{Buffer};
           SS << format("%u", Offset);
@@ -1530,21 +1530,51 @@
   }
 };
 
+class AMDGPULowerModuleLDSLegacy : public ModulePass {
+public:
+  const AMDGPUTargetMachine *TM;
+  static char ID;
+
+  AMDGPULowerModuleLDSLegacy(const AMDGPUTargetMachine *TM_ = nullptr)
+      : ModulePass(ID), TM(TM_) {
+    initializeAMDGPULowerModuleLDSLegacyPass(*PassRegistry::getPassRegistry());
+  }
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    if (!TM)
+      AU.addRequired<TargetPassConfig>();
+  }
+
+  bool runOnModule(Module &M) override {
+    if (!TM) {
+      auto &TPC = getAnalysis<TargetPassConfig>();
+      TM = &TPC.getTM<AMDGPUTargetMachine>();
+    }
+
+    return AMDGPULowerModuleLDS(*TM).runOnModule(M);
+  }
+};
+
 } // namespace
-char AMDGPULowerModuleLDS::ID = 0;
+char AMDGPULowerModuleLDSLegacy::ID = 0;
 
-char &llvm::AMDGPULowerModuleLDSID = AMDGPULowerModuleLDS::ID;
+char &llvm::AMDGPULowerModuleLDSLegacyPassID = AMDGPULowerModuleLDSLegacy::ID;
 
-INITIALIZE_PASS(AMDGPULowerModuleLDS, DEBUG_TYPE,
-                "Lower uses of LDS variables from non-kernel functions", false,
-                false)
+INITIALIZE_PASS_BEGIN(AMDGPULowerModuleLDSLegacy, DEBUG_TYPE,
+                      "Lower uses of LDS variables from non-kernel functions",
+                      false, false)
+INITIALIZE_PASS_DEPENDENCY(TargetPassConfig)
+INITIALIZE_PASS_END(AMDGPULowerModuleLDSLegacy, DEBUG_TYPE,
+                    "Lower uses of LDS variables from non-kernel functions",
+                    false, false)
 
-ModulePass *llvm::createAMDGPULowerModuleLDSPass() {
-  return new AMDGPULowerModuleLDS();
+ModulePass *
+llvm::createAMDGPULowerModuleLDSLegacyPass(const AMDGPUTargetMachine *TM) {
+  return new AMDGPULowerModuleLDSLegacy(TM);
 }
 
 PreservedAnalyses AMDGPULowerModuleLDSPass::run(Module &M,
                                                 ModuleAnalysisManager &) {
-  return AMDGPULowerModuleLDS().runOnModule(M) ? PreservedAnalyses::none()
-                                               : PreservedAnalyses::all();
+  return AMDGPULowerModuleLDS(TM).runOnModule(M) ? PreservedAnalyses::none()
+                                                 : PreservedAnalyses::all();
 }
Index: llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -185,7 +185,7 @@
                       "AMDGPU promote alloca to vector or LDS", false, false)
 // Move LDS uses from functions to kernels before promote alloca for accurate
 // estimation of LDS available
-INITIALIZE_PASS_DEPENDENCY(AMDGPULowerModuleLDS)
+INITIALIZE_PASS_DEPENDENCY(AMDGPULowerModuleLDSLegacy)
 INITIALIZE_PASS_END(AMDGPUPromoteAlloca, DEBUG_TYPE,
                     "AMDGPU promote alloca to vector or LDS", false, false)
 
Index: llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -388,7 +388,7 @@
   initializeAMDGPUCodeGenPreparePass(*PR);
   initializeAMDGPULateCodeGenPreparePass(*PR);
   initializeAMDGPURemoveIncompatibleFunctionsPass(*PR);
-  initializeAMDGPULowerModuleLDSPass(*PR);
+  initializeAMDGPULowerModuleLDSLegacyPass(*PR);
   initializeAMDGPURewriteOutArgumentsPass(*PR);
   initializeAMDGPURewriteUndefForPHIPass(*PR);
   initializeAMDGPUUnifyMetadataPass(*PR);
@@ -595,8 +595,8 @@
 
 void AMDGPUTargetMachine::registerPassBuilderCallbacks(PassBuilder &PB) {
   PB.registerPipelineParsingCallback(
-      [](StringRef PassName, ModulePassManager &PM,
-         ArrayRef<PassBuilder::PipelineElement>) {
+      [this](StringRef PassName, ModulePassManager &PM,
+             ArrayRef<PassBuilder::PipelineElement>) {
         if (PassName == "amdgpu-unify-metadata") {
           PM.addPass(AMDGPUUnifyMetadataPass());
           return true;
@@ -610,7 +610,7 @@
           return true;
         }
         if (PassName == "amdgpu-lower-module-lds") {
-          PM.addPass(AMDGPULowerModuleLDSPass());
+          PM.addPass(AMDGPULowerModuleLDSPass(*this));
           return true;
         }
         if (PassName == "amdgpu-lower-ctor-dtor") {
@@ -989,7 +989,7 @@
 
   // Runs before PromoteAlloca so the latter can account for function uses
   if (EnableLowerModuleLDS) {
-    addPass(createAMDGPULowerModuleLDSPass());
+    addPass(createAMDGPULowerModuleLDSLegacyPass(&TM));
   }
 
   // AMDGPUAttributor infers lack of llvm.amdgcn.lds.kernel.id calls, so run