Reapply: R600: Make sure to inline all internal functions

[oota-llvm.git] / lib / Target / R600 / AMDGPUTargetMachine.cpp

diff --git a/lib/Target/R600/AMDGPUTargetMachine.cpp b/lib/Target/R600/AMDGPUTargetMachine.cpp
index 1b4fe832f20b80d1867a02160a8398233db0b366..ca4d0cae9b25cd70498bb59d47476e5314e8fb14 100644 (file)
--- a/lib/Target/R600/AMDGPUTargetMachine.cpp
+++ b/lib/Target/R600/AMDGPUTargetMachine.cpp
@@ -80,6 +80,7 @@ public:
     return nullptr;
   }
 
+  void addIRPasses() override;
   void addCodeGenPrepare() override;
   bool addPreISel() override;
   bool addInstSelector() override;
@@ -106,6 +107,19 @@ void AMDGPUTargetMachine::addAnalysisPasses(PassManagerBase &PM) {
   PM.add(createAMDGPUTargetTransformInfoPass(this));
 }
 
+void AMDGPUPassConfig::addIRPasses() {
+  // Function calls are not supported, so make sure we inline everything.
+  addPass(createAMDGPUAlwaysInlinePass());
+  addPass(createAlwaysInlinerPass());
+  // We need to add the barrier noop pass, otherwise adding the function
+  // inlining pass will cause all of the PassConfigs passes to be run
+  // one function at a time, which means if we have a nodule with two
+  // functions, then we will generate code for the first function
+  // without ever running any passes on the second.
+  addPass(createBarrierNoopPass());
+  TargetPassConfig::addIRPasses();
+}
+
 void AMDGPUPassConfig::addCodeGenPrepare() {
   const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
   if (ST.isPromoteAllocaEnabled()) {