diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.h b/llvm/lib/Target/AMDGPU/AMDGPU.h
--- a/llvm/lib/Target/AMDGPU/AMDGPU.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.h
@@ -253,6 +253,15 @@
   TargetMachine *TM = nullptr,
   CodeGenOpt::Level OptLevel = CodeGenOpt::Default);
 ModulePass *createAMDGPUAlwaysInlinePass(bool GlobalOpt = true);
+
+struct AMDGPUAlwaysInlinePass : PassInfoMixin<AMDGPUAlwaysInlinePass> {
+  AMDGPUAlwaysInlinePass(bool GlobalOpt = true) : GlobalOpt(GlobalOpt) {}
+  PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
+
+private:
+  bool GlobalOpt;
+};
+
 ModulePass *createR600OpenCLImageTypeLoweringPass();
 FunctionPass *createAMDGPUAnnotateUniformValues();
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
@@ -17,6 +17,7 @@
 #include "Utils/AMDGPUBaseInfo.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/IR/Module.h"
+#include "llvm/IR/PassManager.h"
 #include "llvm/Transforms/Utils/Cloning.h"
 
 using namespace llvm;
@@ -32,8 +33,6 @@
 class AMDGPUAlwaysInline : public ModulePass {
   bool GlobalOpt;
 
-  void recursivelyVisitUsers(GlobalValue &GV,
-                             SmallPtrSetImpl<Function *> &FuncsToAlwaysInline);
 public:
   static char ID;
 
@@ -53,9 +52,9 @@
 
 char AMDGPUAlwaysInline::ID = 0;
 
-void AMDGPUAlwaysInline::recursivelyVisitUsers(
-  GlobalValue &GV,
-  SmallPtrSetImpl<Function *> &FuncsToAlwaysInline) {
+static void
+recursivelyVisitUsers(GlobalValue &GV,
+                      SmallPtrSetImpl<Function *> &FuncsToAlwaysInline) {
   SmallVector<User *, 16> Stack;
 
   SmallPtrSet<const Value *, 8> Visited;
@@ -91,7 +90,7 @@
   }
 }
 
-bool AMDGPUAlwaysInline::runOnModule(Module &M) {
+static bool alwaysInlineImpl(Module &M, bool GlobalOpt) {
   std::vector<GlobalAlias*> AliasesToRemove;
 
   SmallPtrSet<Function *, 8> FuncsToAlwaysInline;
@@ -157,7 +156,16 @@
   return !FuncsToAlwaysInline.empty() || !FuncsToNoInline.empty();
 }
 
+bool AMDGPUAlwaysInline::runOnModule(Module &M) {
+  return alwaysInlineImpl(M, GlobalOpt);
+}
+
 ModulePass *llvm::createAMDGPUAlwaysInlinePass(bool GlobalOpt) {
   return new AMDGPUAlwaysInline(GlobalOpt);
 }
 
+PreservedAnalyses AMDGPUAlwaysInlinePass::run(Module &M,
+                                              ModuleAnalysisManager &AM) {
+  alwaysInlineImpl(M, GlobalOpt);
+  return PreservedAnalyses::all();
+}
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -506,6 +506,10 @@
           PM.addPass(AMDGPUPrintfRuntimeBindingPass());
           return true;
         }
+        if (PassName == "amdgpu-always-inline") {
+          PM.addPass(AMDGPUAlwaysInlinePass());
+          return true;
+        }
         return false;
       });
   PB.registerPipelineParsingCallback(
@@ -565,6 +569,8 @@
         if (InternalizeSymbols) {
           PM.addPass(GlobalDCEPass());
         }
+        if (EarlyInlineAll && !EnableFunctionCalls)
+          PM.addPass(AMDGPUAlwaysInlinePass());
       });
 
   PB.registerCGSCCOptimizerLateEPCallback(
diff --git a/llvm/test/CodeGen/AMDGPU/force-alwaysinline-lds-global-address.ll b/llvm/test/CodeGen/AMDGPU/force-alwaysinline-lds-global-address.ll
--- a/llvm/test/CodeGen/AMDGPU/force-alwaysinline-lds-global-address.ll
+++ b/llvm/test/CodeGen/AMDGPU/force-alwaysinline-lds-global-address.ll
@@ -1,5 +1,7 @@
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-always-inline %s | FileCheck -check-prefixes=CALLS-ENABLED,ALL %s
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=amdgpu-always-inline %s | FileCheck -check-prefixes=CALLS-ENABLED,ALL %s
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-stress-function-calls -amdgpu-always-inline %s | FileCheck -check-prefixes=STRESS-CALLS,ALL %s
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-stress-function-calls -passes=amdgpu-always-inline %s | FileCheck -check-prefixes=STRESS-CALLS,ALL %s
 
 target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5"
 
diff --git a/llvm/tools/opt/opt.cpp b/llvm/tools/opt/opt.cpp
--- a/llvm/tools/opt/opt.cpp
+++ b/llvm/tools/opt/opt.cpp
@@ -471,7 +471,8 @@
       "amdgpu-propagate-attributes-early",
       "amdgpu-propagate-attributes-late",
       "amdgpu-unify-metadata",
-      "amdgpu-printf-runtime-binding"};
+      "amdgpu-printf-runtime-binding",
+      "amdgpu-always-inline"};
   for (const auto &P : PassNameExactToIgnore)
     if (Pass == P)
       return false;