Loop strength reduce: fix function name.

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

diff --git a/lib/Target/R600/AMDGPUTargetMachine.cpp b/lib/Target/R600/AMDGPUTargetMachine.cpp
index 31fbf32d0c9a60d9c329e9c351a07af38bf59b77..a550cc0178ae7e0e5c0d197350e2d5903846a17f 100644 (file)
--- a/lib/Target/R600/AMDGPUTargetMachine.cpp
+++ b/lib/Target/R600/AMDGPUTargetMachine.cpp
@@ -21,10 +21,10 @@
 #include "SIISelLowering.h"
 #include "SIInstrInfo.h"
 #include "llvm/Analysis/Passes.h"
-#include "llvm/Analysis/Verifier.h"
 #include "llvm/CodeGen/MachineFunctionAnalysis.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
 #include "llvm/CodeGen/Passes.h"
+#include "llvm/IR/Verifier.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/PassManager.h"
 #include "llvm/Support/TargetRegistry.h"
@@ -33,6 +33,7 @@
 #include "llvm/Transforms/Scalar.h"
 #include <llvm/CodeGen/Passes.h>
 
+
 using namespace llvm;
 
 extern "C" void LLVMInitializeR600Target() {
@@ -41,13 +42,27 @@ extern "C" void LLVMInitializeR600Target() {
 }
 
 static ScheduleDAGInstrs *createR600MachineScheduler(MachineSchedContext *C) {
-  return new ScheduleDAGMI(C, new R600SchedStrategy());
+  return new ScheduleDAGMILive(C, new R600SchedStrategy());
 }
 
 static MachineSchedRegistry
 SchedCustomRegistry("r600", "Run R600's custom scheduler",
                     createR600MachineScheduler);
 
+static std::string computeDataLayout(const AMDGPUSubtarget &ST) {
+  std::string Ret = "e-p:32:32";
+
+  if (ST.is64bit()) {
+    // 32-bit private, local, and region pointers. 64-bit global and constant.
+    Ret += "-p1:64:64-p2:64:64-p3:32:32-p4:32:32-p5:64:64";
+  }
+
+  Ret += "-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256"
+         "-v512:512-v1024:1024-v2048:2048-n32:64";
+
+  return Ret;
+}
+
 AMDGPUTargetMachine::AMDGPUTargetMachine(const Target &T, StringRef TT,
     StringRef CPU, StringRef FS,
   TargetOptions Options,
@@ -57,19 +72,22 @@ AMDGPUTargetMachine::AMDGPUTargetMachine(const Target &T, StringRef TT,
 :
   LLVMTargetMachine(T, TT, CPU, FS, Options, RM, CM, OptLevel),
   Subtarget(TT, CPU, FS),
-  Layout(Subtarget.getDataLayout()),
+  Layout(computeDataLayout(Subtarget)),
   FrameLowering(TargetFrameLowering::StackGrowsUp,
-      Subtarget.device()->getStackAlignment(), 0),
+                64 * 16 // Maximum stack alignment (long16)
+               , 0),
   IntrinsicInfo(this),
   InstrItins(&Subtarget.getInstrItineraryData()) {
   // TLInfo uses InstrInfo so it must be initialized after.
-  if (Subtarget.device()->getGeneration() <= AMDGPUDeviceInfo::HD6XXX) {
-    InstrInfo = new R600InstrInfo(*this);
-    TLInfo = new R600TargetLowering(*this);
+  if (Subtarget.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS) {
+    InstrInfo.reset(new R600InstrInfo(*this));
+    TLInfo.reset(new R600TargetLowering(*this));
   } else {
-    InstrInfo = new SIInstrInfo(*this);
-    TLInfo = new SITargetLowering(*this);
+    InstrInfo.reset(new SIInstrInfo(*this));
+    TLInfo.reset(new SITargetLowering(*this));
   }
+  setRequiresStructuredCFG(true);
+  initAsmInfo();
 }
 
 AMDGPUTargetMachine::~AMDGPUTargetMachine() {
@@ -79,18 +97,20 @@ namespace {
 class AMDGPUPassConfig : public TargetPassConfig {
 public:
   AMDGPUPassConfig(AMDGPUTargetMachine *TM, PassManagerBase &PM)
-    : TargetPassConfig(TM, PM) {
-    const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
-    if (ST.device()->getGeneration() <= AMDGPUDeviceInfo::HD6XXX) {
-      enablePass(&MachineSchedulerID);
-      MachineSchedRegistry::setDefault(createR600MachineScheduler);
-    }
-  }
+    : TargetPassConfig(TM, PM) {}
 
   AMDGPUTargetMachine &getAMDGPUTargetMachine() const {
     return getTM<AMDGPUTargetMachine>();
   }
 
+  virtual ScheduleDAGInstrs *
+  createMachineScheduler(MachineSchedContext *C) const {
+    const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
+    if (ST.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS)
+      return createR600MachineScheduler(C);
+    return 0;
+  }
+
   virtual bool addPreISel();
   virtual bool addInstSelector();
   virtual bool addPreRegAlloc();
@@ -104,53 +124,76 @@ TargetPassConfig *AMDGPUTargetMachine::createPassConfig(PassManagerBase &PM) {
   return new AMDGPUPassConfig(this, PM);
 }
 
+//===----------------------------------------------------------------------===//
+// AMDGPU Analysis Pass Setup
+//===----------------------------------------------------------------------===//
+
+void AMDGPUTargetMachine::addAnalysisPasses(PassManagerBase &PM) {
+  // Add first the target-independent BasicTTI pass, then our AMDGPU pass. This
+  // allows the AMDGPU pass to delegate to the target independent layer when
+  // appropriate.
+  PM.add(createBasicTargetTransformInfoPass(this));
+  PM.add(createAMDGPUTargetTransformInfoPass(this));
+}
+
 bool
 AMDGPUPassConfig::addPreISel() {
   const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
-  if (ST.device()->getGeneration() > AMDGPUDeviceInfo::HD6XXX) {
-    addPass(createAMDGPUStructurizeCFGPass());
+  addPass(createFlattenCFGPass());
+  if (ST.IsIRStructurizerEnabled())
+    addPass(createStructurizeCFGPass());
+  if (ST.getGeneration() > AMDGPUSubtarget::NORTHERN_ISLANDS) {
+    addPass(createSinkingPass());
+    addPass(createSITypeRewriter());
     addPass(createSIAnnotateControlFlowPass());
+  } else {
+    addPass(createR600TextureIntrinsicsReplacer());
   }
   return false;
 }
 
 bool AMDGPUPassConfig::addInstSelector() {
   addPass(createAMDGPUISelDag(getAMDGPUTargetMachine()));
-
-  const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
-  if (ST.device()->getGeneration() <= AMDGPUDeviceInfo::HD6XXX) {
-    // This callbacks this pass uses are not implemented yet on SI.
-    addPass(createAMDGPUIndirectAddressingPass(*TM));
-  }
   return false;
 }
 
 bool AMDGPUPassConfig::addPreRegAlloc() {
   addPass(createAMDGPUConvertToISAPass(*TM));
+  const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
+
+  if (ST.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS) {
+    addPass(createR600VectorRegMerger(*TM));
+  } else {
+    addPass(createSIFixSGPRCopiesPass(*TM));
+  }
   return false;
 }
 
 bool AMDGPUPassConfig::addPostRegAlloc() {
   const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
 
-  if (ST.device()->getGeneration() > AMDGPUDeviceInfo::HD6XXX) {
+  if (ST.getGeneration() > AMDGPUSubtarget::NORTHERN_ISLANDS) {
     addPass(createSIInsertWaits(*TM));
   }
   return false;
 }
 
 bool AMDGPUPassConfig::addPreSched2() {
+  const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
 
-  addPass(&IfConverterID);
+  if (ST.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS)
+    addPass(createR600EmitClauseMarkers());
+  if (ST.isIfCvtEnabled())
+    addPass(&IfConverterID);
+  if (ST.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS)
+    addPass(createR600ClauseMergePass(*TM));
   return false;
 }
 
 bool AMDGPUPassConfig::addPreEmitPass() {
   const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>();
-  if (ST.device()->getGeneration() <= AMDGPUDeviceInfo::HD6XXX) {
-    addPass(createAMDGPUCFGPreparationPass(*TM));
-    addPass(createAMDGPUCFGStructurizerPass(*TM));
-    addPass(createR600EmitClauseMarkers(*TM));
+  if (ST.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS) {
+    addPass(createAMDGPUCFGStructurizerPass());
     addPass(createR600ExpandSpecialInstrsPass(*TM));
     addPass(&FinalizeMachineBundlesID);
     addPass(createR600Packetizer(*TM));
@@ -161,4 +204,3 @@ bool AMDGPUPassConfig::addPreEmitPass() {
 
   return false;
 }
-