summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--lib/Target/AMDGPU/AMDGPUISelLowering.h9
-rw-r--r--lib/Target/AMDGPU/AMDGPUInstrInfo.h6
-rw-r--r--lib/Target/AMDGPU/AMDGPURegisterInfo.h9
-rw-r--r--lib/Target/AMDGPU/AMDGPUSubtarget.cpp28
-rw-r--r--lib/Target/AMDGPU/AMDGPUSubtarget.h28
-rw-r--r--lib/Target/AMDGPU/AMDGPUTargetMachine.cpp5
-rw-r--r--lib/Target/AMDGPU/AMDGPUTargetMachine.h1
-rw-r--r--lib/Target/AMDGPU/AMDIL.h6
-rw-r--r--lib/Target/AMDGPU/AMDIL7XXDevice.cpp6
-rw-r--r--lib/Target/AMDGPU/AMDIL7XXDevice.h6
-rw-r--r--lib/Target/AMDGPU/AMDILBase.td8
-rw-r--r--lib/Target/AMDGPU/AMDILCFGStructurizer.cpp56
-rw-r--r--lib/Target/AMDGPU/AMDILDevice.cpp4
-rw-r--r--lib/Target/AMDGPU/AMDILDevice.h8
-rw-r--r--lib/Target/AMDGPU/AMDILDeviceInfo.h12
-rw-r--r--lib/Target/AMDGPU/AMDILDevices.h6
-rw-r--r--lib/Target/AMDGPU/AMDILEvergreenDevice.cpp8
-rw-r--r--lib/Target/AMDGPU/AMDILEvergreenDevice.h6
-rw-r--r--lib/Target/AMDGPU/AMDILFrameLowering.h6
-rw-r--r--lib/Target/AMDGPU/AMDILISelDAGToDAG.cpp172
-rw-r--r--lib/Target/AMDGPU/AMDILISelLowering.cpp33
-rw-r--r--lib/Target/AMDGPU/AMDILIntrinsicInfo.h6
-rw-r--r--lib/Target/AMDGPU/AMDILNIDevice.cpp8
-rw-r--r--lib/Target/AMDGPU/AMDILNIDevice.h6
-rw-r--r--lib/Target/AMDGPU/AMDILPeepholeOptimizer.cpp3
-rw-r--r--lib/Target/AMDGPU/AMDILRegisterInfo.td9
-rw-r--r--lib/Target/AMDGPU/AMDILSIDevice.h6
-rw-r--r--lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.cpp9
-rw-r--r--lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.h6
-rw-r--r--lib/Target/AMDGPU/MCTargetDesc/R600MCCodeEmitter.cpp45
-rw-r--r--lib/Target/AMDGPU/R600ExpandSpecialInstrs.cpp6
-rw-r--r--lib/Target/AMDGPU/R600ISelLowering.cpp336
-rw-r--r--lib/Target/AMDGPU/R600ISelLowering.h3
-rw-r--r--lib/Target/AMDGPU/R600InstrInfo.cpp3
-rw-r--r--lib/Target/AMDGPU/R600Instructions.td38
-rw-r--r--lib/Target/AMDGPU/R600MachineFunctionInfo.cpp3
-rw-r--r--lib/Target/AMDGPU/R600RegisterInfo.h3
-rw-r--r--lib/Target/AMDGPU/SIFixSGPRLiveness.cpp3
-rw-r--r--lib/Target/AMDGPU/SIISelLowering.cpp10
-rw-r--r--lib/Target/AMDGPU/SIISelLowering.h3
-rw-r--r--lib/Target/AMDGPU/SIInstrInfo.cpp2
-rw-r--r--lib/Target/AMDGPU/SIInstrInfo.td2
-rw-r--r--lib/Target/AMDGPU/SIInstructions.td24
-rw-r--r--lib/Target/AMDGPU/SILowerControlFlow.cpp6
-rw-r--r--lib/Target/AMDGPU/SIMachineFunctionInfo.h4
-rw-r--r--lib/Target/AMDGPU/SIRegisterInfo.h3
46 files changed, 452 insertions, 518 deletions
diff --git a/lib/Target/AMDGPU/AMDGPUISelLowering.h b/lib/Target/AMDGPU/AMDGPUISelLowering.h
index 58d2287a34..7783326d03 100644
--- a/lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ b/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -21,8 +21,7 @@ namespace llvm {
class MachineRegisterInfo;
-class AMDGPUTargetLowering : public TargetLowering
-{
+class AMDGPUTargetLowering : public TargetLowering {
private:
SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerUDIVREM(SDValue Op, SelectionDAG &DAG) const;
@@ -96,11 +95,9 @@ private:
SDValue LowerFP_ROUND(SDValue Op, SelectionDAG &DAG) const;
};
-namespace AMDGPUISD
-{
+namespace AMDGPUISD {
-enum
-{
+enum {
// AMDIL ISD Opcodes
FIRST_NUMBER = ISD::BUILTIN_OP_END,
MAD, // 32bit Fused Multiply Add instruction
diff --git a/lib/Target/AMDGPU/AMDGPUInstrInfo.h b/lib/Target/AMDGPU/AMDGPUInstrInfo.h
index e81f73d8f5..bd6a0283cf 100644
--- a/lib/Target/AMDGPU/AMDGPUInstrInfo.h
+++ b/lib/Target/AMDGPU/AMDGPUInstrInfo.h
@@ -12,8 +12,8 @@
//
//===----------------------------------------------------------------------===//
-#ifndef AMDGPUINSTRUCTIONINFO_H_
-#define AMDGPUINSTRUCTIONINFO_H_
+#ifndef AMDGPUINSTRUCTIONINFO_H
+#define AMDGPUINSTRUCTIONINFO_H
#include "AMDGPURegisterInfo.h"
#include "AMDGPUInstrInfo.h"
@@ -145,4 +145,4 @@ public:
} // End llvm namespace
-#endif // AMDGPUINSTRINFO_H_
+#endif // AMDGPUINSTRINFO_H
diff --git a/lib/Target/AMDGPU/AMDGPURegisterInfo.h b/lib/Target/AMDGPU/AMDGPURegisterInfo.h
index 326610d333..14d618b31a 100644
--- a/lib/Target/AMDGPU/AMDGPURegisterInfo.h
+++ b/lib/Target/AMDGPU/AMDGPURegisterInfo.h
@@ -12,8 +12,8 @@
//
//===----------------------------------------------------------------------===//
-#ifndef AMDGPUREGISTERINFO_H_
-#define AMDGPUREGISTERINFO_H_
+#ifndef AMDGPUREGISTERINFO_H
+#define AMDGPUREGISTERINFO_H
#include "llvm/ADT/BitVector.h"
#include "llvm/Target/TargetRegisterInfo.h"
@@ -27,8 +27,7 @@ namespace llvm {
class AMDGPUTargetMachine;
class TargetInstrInfo;
-struct AMDGPURegisterInfo : public AMDGPUGenRegisterInfo
-{
+struct AMDGPURegisterInfo : public AMDGPUGenRegisterInfo {
TargetMachine &TM;
const TargetInstrInfo &TII;
static const uint16_t CalleeSavedReg;
@@ -59,4 +58,4 @@ struct AMDGPURegisterInfo : public AMDGPUGenRegisterInfo
} // End namespace llvm
-#endif // AMDIDSAREGISTERINFO_H_
+#endif // AMDIDSAREGISTERINFO_H
diff --git a/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
index cc8f961853..461a22fff0 100644
--- a/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ b/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -21,24 +21,24 @@ using namespace llvm;
#include "AMDGPUGenSubtargetInfo.inc"
AMDGPUSubtarget::AMDGPUSubtarget(StringRef TT, StringRef CPU, StringRef FS) :
- AMDGPUGenSubtargetInfo(TT, CPU, FS), mDumpCode(false) {
+ AMDGPUGenSubtargetInfo(TT, CPU, FS), DumpCode(false) {
InstrItins = getInstrItineraryForCPU(CPU);
memset(CapsOverride, 0, sizeof(*CapsOverride)
* AMDGPUDeviceInfo::MaxNumberCapabilities);
// Default card
StringRef GPU = CPU;
- mIs64bit = false;
- mDefaultSize[0] = 64;
- mDefaultSize[1] = 1;
- mDefaultSize[2] = 1;
+ Is64bit = false;
+ DefaultSize[0] = 64;
+ DefaultSize[1] = 1;
+ DefaultSize[2] = 1;
ParseSubtargetFeatures(GPU, FS);
- mDevName = GPU;
- mDevice = AMDGPUDeviceInfo::getDeviceFromName(mDevName, this, mIs64bit);
+ DevName = GPU;
+ Device = AMDGPUDeviceInfo::getDeviceFromName(DevName, this, Is64bit);
}
AMDGPUSubtarget::~AMDGPUSubtarget() {
- delete mDevice;
+ delete Device;
}
bool
@@ -49,7 +49,7 @@ AMDGPUSubtarget::isOverride(AMDGPUDeviceInfo::Caps caps) const {
}
bool
AMDGPUSubtarget::is64bit() const {
- return mIs64bit;
+ return Is64bit;
}
bool
AMDGPUSubtarget::isTargetELF() const {
@@ -60,27 +60,27 @@ AMDGPUSubtarget::getDefaultSize(uint32_t dim) const {
if (dim > 3) {
return 1;
} else {
- return mDefaultSize[dim];
+ return DefaultSize[dim];
}
}
std::string
AMDGPUSubtarget::getDataLayout() const {
- if (!mDevice) {
+ if (!Device) {
return std::string("e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16"
"-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f80:32:32"
"-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64"
"-v96:128:128-v128:128:128-v192:256:256-v256:256:256"
"-v512:512:512-v1024:1024:1024-v2048:2048:2048-a0:0:64");
}
- return mDevice->getDataLayout();
+ return Device->getDataLayout();
}
std::string
AMDGPUSubtarget::getDeviceName() const {
- return mDevName;
+ return DevName;
}
const AMDGPUDevice *
AMDGPUSubtarget::device() const {
- return mDevice;
+ return Device;
}
diff --git a/lib/Target/AMDGPU/AMDGPUSubtarget.h b/lib/Target/AMDGPU/AMDGPUSubtarget.h
index 30bda83a20..79a74865f6 100644
--- a/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -11,8 +11,8 @@
//
//===----------------------------------------------------------------------===//
-#ifndef _AMDGPUSUBTARGET_H_
-#define _AMDGPUSUBTARGET_H_
+#ifndef AMDGPUSUBTARGET_H
+#define AMDGPUSUBTARGET_H
#include "AMDILDevice.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/ADT/StringRef.h"
@@ -25,18 +25,16 @@
namespace llvm {
-class AMDGPUSubtarget : public AMDGPUGenSubtargetInfo
-{
+class AMDGPUSubtarget : public AMDGPUGenSubtargetInfo {
private:
bool CapsOverride[AMDGPUDeviceInfo::MaxNumberCapabilities];
- const AMDGPUDevice *mDevice;
- size_t mDefaultSize[3];
- size_t mMinimumSize[3];
- std::string mDevName;
- bool mIs64bit;
- bool mIs32on64bit;
- bool mDumpCode;
- bool mR600ALUInst;
+ const AMDGPUDevice *Device;
+ size_t DefaultSize[3];
+ std::string DevName;
+ bool Is64bit;
+ bool Is32on64bit;
+ bool DumpCode;
+ bool R600ALUInst;
InstrItineraryData InstrItins;
@@ -56,11 +54,11 @@ public:
std::string getDataLayout() const;
std::string getDeviceName() const;
virtual size_t getDefaultSize(uint32_t dim) const;
- bool dumpCode() const { return mDumpCode; }
- bool r600ALUEncoding() const { return mR600ALUInst; }
+ bool dumpCode() const { return DumpCode; }
+ bool r600ALUEncoding() const { return R600ALUInst; }
};
} // End namespace llvm
-#endif // AMDGPUSUBTARGET_H_
+#endif // AMDGPUSUBTARGET_H
diff --git a/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 113b7bfb3a..50a3f62b27 100644
--- a/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -51,10 +51,7 @@ AMDGPUTargetMachine::AMDGPUTargetMachine(const Target &T, StringRef TT,
FrameLowering(TargetFrameLowering::StackGrowsUp,
Subtarget.device()->getStackAlignment(), 0),
IntrinsicInfo(this),
- InstrItins(&Subtarget.getInstrItineraryData()),
- mDump(false)
-
-{
+ InstrItins(&Subtarget.getInstrItineraryData()) {
// TLInfo uses InstrInfo so it must be initialized after.
if (Subtarget.device()->getGeneration() <= AMDGPUDeviceInfo::HD6XXX) {
InstrInfo = new R600InstrInfo(*this);
diff --git a/lib/Target/AMDGPU/AMDGPUTargetMachine.h b/lib/Target/AMDGPU/AMDGPUTargetMachine.h
index 712bc6c3dc..7924f10974 100644
--- a/lib/Target/AMDGPU/AMDGPUTargetMachine.h
+++ b/lib/Target/AMDGPU/AMDGPUTargetMachine.h
@@ -35,7 +35,6 @@ class AMDGPUTargetMachine : public LLVMTargetMachine {
const AMDGPUInstrInfo * InstrInfo;
AMDGPUTargetLowering * TLInfo;
const InstrItineraryData* InstrItins;
- bool mDump;
public:
AMDGPUTargetMachine(const Target &T, StringRef TT, StringRef FS,
diff --git a/lib/Target/AMDGPU/AMDIL.h b/lib/Target/AMDGPU/AMDIL.h
index e96b123bb7..d8682fb727 100644
--- a/lib/Target/AMDGPU/AMDIL.h
+++ b/lib/Target/AMDGPU/AMDIL.h
@@ -12,8 +12,8 @@
//
//===----------------------------------------------------------------------===//
-#ifndef AMDIL_H_
-#define AMDIL_H_
+#ifndef AMDIL_H
+#define AMDIL_H
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/Target/TargetMachine.h"
@@ -103,4 +103,4 @@ enum AddressSpaces {
} // namespace AMDGPUAS
} // end namespace llvm
-#endif // AMDIL_H_
+#endif // AMDIL_H
diff --git a/lib/Target/AMDGPU/AMDIL7XXDevice.cpp b/lib/Target/AMDGPU/AMDIL7XXDevice.cpp
index 854d6901d7..21879d6cae 100644
--- a/lib/Target/AMDGPU/AMDIL7XXDevice.cpp
+++ b/lib/Target/AMDGPU/AMDIL7XXDevice.cpp
@@ -16,11 +16,11 @@ AMDGPU7XXDevice::AMDGPU7XXDevice(AMDGPUSubtarget *ST) : AMDGPUDevice(ST) {
setCaps();
std::string name = mSTM->getDeviceName();
if (name == "rv710") {
- mDeviceFlag = OCL_DEVICE_RV710;
+ DeviceFlag = OCL_DEVICE_RV710;
} else if (name == "rv730") {
- mDeviceFlag = OCL_DEVICE_RV730;
+ DeviceFlag = OCL_DEVICE_RV730;
} else {
- mDeviceFlag = OCL_DEVICE_RV770;
+ DeviceFlag = OCL_DEVICE_RV770;
}
}
diff --git a/lib/Target/AMDGPU/AMDIL7XXDevice.h b/lib/Target/AMDGPU/AMDIL7XXDevice.h
index e848e2e0f2..1befb92485 100644
--- a/lib/Target/AMDGPU/AMDIL7XXDevice.h
+++ b/lib/Target/AMDGPU/AMDIL7XXDevice.h
@@ -14,8 +14,8 @@
// implement in order to correctly answer queries on the capabilities of the
// specific hardware.
//===----------------------------------------------------------------------===//
-#ifndef _AMDIL7XXDEVICEIMPL_H_
-#define _AMDIL7XXDEVICEIMPL_H_
+#ifndef AMDIL7XXDEVICEIMPL_H
+#define AMDIL7XXDEVICEIMPL_H
#include "AMDILDevice.h"
namespace llvm {
@@ -67,4 +67,4 @@ public:
}; // AMDGPU710Device
} // namespace llvm
-#endif // _AMDILDEVICEIMPL_H_
+#endif // AMDILDEVICEIMPL_H
diff --git a/lib/Target/AMDGPU/AMDILBase.td b/lib/Target/AMDGPU/AMDILBase.td
index ffe9ce2c53..c12cedcf7f 100644
--- a/lib/Target/AMDGPU/AMDILBase.td
+++ b/lib/Target/AMDGPU/AMDILBase.td
@@ -52,12 +52,12 @@ def FeatureNoInline : SubtargetFeature<"no-inline",
"specify whether to not inline functions">;
def Feature64BitPtr : SubtargetFeature<"64BitPtr",
- "mIs64bit",
+ "Is64bit",
"false",
"Specify if 64bit addressing should be used.">;
def Feature32on64BitPtr : SubtargetFeature<"64on32BitPtr",
- "mIs32on64bit",
+ "Is32on64bit",
"false",
"Specify if 64bit sized pointers with 32bit addressing should be used.">;
def FeatureDebug : SubtargetFeature<"debug",
@@ -65,12 +65,12 @@ def FeatureDebug : SubtargetFeature<"debug",
"true",
"Debug mode is enabled, so disable hardware accelerated address spaces.">;
def FeatureDumpCode : SubtargetFeature <"DumpCode",
- "mDumpCode",
+ "DumpCode",
"true",
"Dump MachineInstrs in the CodeEmitter">;
def FeatureR600ALUInst : SubtargetFeature<"R600ALUInst",
- "mR600ALUInst",
+ "R600ALUInst",
"false",
"Older version of ALU instructions encoding.">;
diff --git a/lib/Target/AMDGPU/AMDILCFGStructurizer.cpp b/lib/Target/AMDGPU/AMDILCFGStructurizer.cpp
index 61d4c5929b..948bdb40d9 100644
--- a/lib/Target/AMDGPU/AMDILCFGStructurizer.cpp
+++ b/lib/Target/AMDGPU/AMDILCFGStructurizer.cpp
@@ -607,7 +607,7 @@ template<class PassT> void CFGStructurizer<PassT>::orderBlocks() {
if (sccNum == INVALIDSCCNUM) {
errs() << "unreachable block BB" << bb->getNumber() << "\n";
}
- } //end of for
+ }
} //orderBlocks
template<class PassT> int CFGStructurizer<PassT>::patternMatch(BlockT *curBlk) {
@@ -839,8 +839,7 @@ int CFGStructurizer<PassT>::loopbreakPatternMatch(LoopT *loopRep,
int numCloned = 0;
int numSerial = 0;
- if (exitBlkSet.size() == 1)
- {
+ if (exitBlkSet.size() == 1) {
exitLandBlk = *exitBlkSet.begin();
} else {
exitLandBlk = findNearestCommonPostDom(exitBlkSet);
@@ -1396,7 +1395,7 @@ void CFGStructurizer<PassT>::mergeIfthenelseBlock(InstrT *branchInstr,
CFGTraits::insertCondBranchBefore(branchInstrPos,
CFGTraits::getBranchNzeroOpcode(oldOpcode),
passRep,
- branchDL);
+ branchDL);
if (trueBlk) {
curBlk->splice(branchInstrPos, trueBlk, trueBlk->begin(), trueBlk->end());
@@ -1455,8 +1454,7 @@ void CFGStructurizer<PassT>::mergeLooplandBlock(BlockT *dstBlk,
// Loop breakInitRegs are init before entering the loop.
for (typename std::set<RegiT>::const_iterator iter =
loopLand->breakInitRegs.begin(),
- iterEnd = loopLand->breakInitRegs.end(); iter != iterEnd; ++iter)
- {
+ iterEnd = loopLand->breakInitRegs.end(); iter != iterEnd; ++iter) {
CFGTraits::insertAssignInstrBefore(dstBlk, passRep, *iter, 0);
}
// Loop endbranchInitRegs are init before entering the loop.
@@ -1608,7 +1606,7 @@ void CFGStructurizer<PassT>::settleLoopcontBlock(BlockT *contingBlk,
CFGTraits::getInstrPos(contingBlk, branchInstr);
BlockT *trueBranch = CFGTraits::getTrueBranch(branchInstr);
int oldOpcode = branchInstr->getOpcode();
- DebugLoc DL = branchInstr->getDebugLoc();
+ DebugLoc DL = branchInstr->getDebugLoc();
// transform contingBlk to
// if () {
@@ -1623,8 +1621,7 @@ void CFGStructurizer<PassT>::settleLoopcontBlock(BlockT *contingBlk,
bool useContinueLogical =
(setReg == INVALIDREGNUM && (&*contingBlk->rbegin()) == branchInstr);
- if (useContinueLogical == false)
- {
+ if (useContinueLogical == false) {
int branchOpcode =
trueBranch == contBlk ? CFGTraits::getBranchNzeroOpcode(oldOpcode)
: CFGTraits::getBranchZeroOpcode(oldOpcode);
@@ -2325,8 +2322,7 @@ void CFGStructurizer<PassT>::addLoopEndbranchInitReg(LoopT *loopRep,
}
theEntry->endbranchInitRegs.insert(regNum);
- if (DEBUGME)
- {
+ if (DEBUGME) {
errs() << "addLoopEndbranchInitReg loop-header = BB"
<< loopRep->getHeader()->getNumber()
<< " regNum = " << regNum << "\n";
@@ -2485,14 +2481,12 @@ public:
private:
-}; //end of class AMDGPUCFGStructurizer
+};
} //end of namespace llvm
-AMDGPUCFGStructurizer::AMDGPUCFGStructurizer(char &pid, TargetMachine &tm
- )
+AMDGPUCFGStructurizer::AMDGPUCFGStructurizer(char &pid, TargetMachine &tm)
: MachineFunctionPass(pid), TM(tm), TII(tm.getInstrInfo()),
- TRI(static_cast<const AMDGPURegisterInfo *>(tm.getRegisterInfo())
- ) {
+ TRI(static_cast<const AMDGPURegisterInfo *>(tm.getRegisterInfo())) {
}
const TargetInstrInfo *AMDGPUCFGStructurizer::getTargetInstrInfo() const {
@@ -2522,7 +2516,7 @@ public:
private:
-}; //end of class AMDGPUCFGPrepare
+};
char AMDGPUCFGPrepare::ID = 0;
} //end of namespace llvm
@@ -2564,7 +2558,7 @@ public:
private:
-}; //end of class AMDGPUCFGPerform
+};
char AMDGPUCFGPerform::ID = 0;
} //end of namespace llvm
@@ -2605,7 +2599,7 @@ struct CFGStructTraits<AMDGPUCFGStructurizer> {
case AMDGPU::SI_IF_NZ: return AMDGPU::SI_IF_NZ;
default:
assert(0 && "internal error");
- };
+ }
return -1;
}
@@ -2617,12 +2611,11 @@ struct CFGStructTraits<AMDGPUCFGStructurizer> {
case AMDGPU::SI_IF_Z: return AMDGPU::SI_IF_Z;
default:
assert(0 && "internal error");
- };
+ }
return -1;
}
- static int getContinueNzeroOpcode(int oldOpcode)
- {
+ static int getContinueNzeroOpcode(int oldOpcode) {
switch(oldOpcode) {
case AMDGPU::JUMP: return AMDGPU::CONTINUE_LOGICALNZ_i32;
default:
@@ -2636,7 +2629,7 @@ struct CFGStructTraits<AMDGPUCFGStructurizer> {
case AMDGPU::JUMP: return AMDGPU::CONTINUE_LOGICALZ_i32;
default:
assert(0 && "internal error");
- };
+ }
return -1;
}
@@ -2689,11 +2682,11 @@ struct CFGStructTraits<AMDGPUCFGStructurizer> {
static DebugLoc getLastDebugLocInBB(MachineBasicBlock *blk) {
//get DebugLoc from the first MachineBasicBlock instruction with debug info
DebugLoc DL;
- for (MachineBasicBlock::iterator iter = blk->begin(); iter != blk->end(); ++iter) {
- MachineInstr *instr = &(*iter);
- if (instr->getDebugLoc().isUnknown() == false) {
- DL = instr->getDebugLoc();
- }
+ for (MachineBasicBlock::iterator iter = blk->begin(); iter != blk->end(); ++iter) {
+ MachineInstr *instr = &(*iter);
+ if (instr->getDebugLoc().isUnknown() == false) {
+ DL = instr->getDebugLoc();
+ }
}
return DL;
}
@@ -2851,7 +2844,7 @@ struct CFGStructTraits<AMDGPUCFGStructurizer> {
static void insertCondBranchBefore(MachineBasicBlock::iterator instrPos,
int newOpcode,
AMDGPUCFGStructurizer *passRep,
- DebugLoc DL) {
+ DebugLoc DL) {
MachineInstr *oldInstr = &(*instrPos);
const TargetInstrInfo *tii = passRep->getTargetInstrInfo();
MachineBasicBlock *blk = oldInstr->getParent();
@@ -2872,7 +2865,7 @@ struct CFGStructTraits<AMDGPUCFGStructurizer> {
int newOpcode,
AMDGPUCFGStructurizer *passRep,
RegiT regNum,
- DebugLoc DL) {
+ DebugLoc DL) {
const TargetInstrInfo *tii = passRep->getTargetInstrInfo();
MachineInstr *newInstr =
@@ -3054,6 +3047,3 @@ bool AMDGPUCFGPerform::runOnMachineFunction(MachineFunction &func) {
*this,
TRI);
}
-
-//end of file newline goes below
-
diff --git a/lib/Target/AMDGPU/AMDILDevice.cpp b/lib/Target/AMDGPU/AMDILDevice.cpp
index 3ff62a8070..393868189e 100644
--- a/lib/Target/AMDGPU/AMDILDevice.cpp
+++ b/lib/Target/AMDGPU/AMDILDevice.cpp
@@ -15,7 +15,7 @@ AMDGPUDevice::AMDGPUDevice(AMDGPUSubtarget *ST) : mSTM(ST) {
mHWBits.resize(AMDGPUDeviceInfo::MaxNumberCapabilities);
mSWBits.resize(AMDGPUDeviceInfo::MaxNumberCapabilities);
setCaps();
- mDeviceFlag = OCL_DEVICE_ALL;
+ DeviceFlag = OCL_DEVICE_ALL;
}
AMDGPUDevice::~AMDGPUDevice() {
@@ -29,7 +29,7 @@ size_t AMDGPUDevice::getMaxGDSSize() const {
uint32_t
AMDGPUDevice::getDeviceFlag() const {
- return mDeviceFlag;
+ return DeviceFlag;
}
size_t AMDGPUDevice::getMaxNumCBs() const {
diff --git a/lib/Target/AMDGPU/AMDILDevice.h b/lib/Target/AMDGPU/AMDILDevice.h
index 05fe153403..3740e3b165 100644
--- a/lib/Target/AMDGPU/AMDILDevice.h
+++ b/lib/Target/AMDGPU/AMDILDevice.h
@@ -14,8 +14,8 @@
// implement in order to correctly answer queries on the capabilities of the
// specific hardware.
//===----------------------------------------------------------------------===//
-#ifndef _AMDILDEVICEIMPL_H_
-#define _AMDILDEVICEIMPL_H_
+#ifndef AMDILDEVICEIMPL_H
+#define AMDILDEVICEIMPL_H
#include "AMDIL.h"
#include "llvm/ADT/BitVector.h"
@@ -106,11 +106,11 @@ protected:
llvm::BitVector mHWBits;
llvm::BitVector mSWBits;
AMDGPUSubtarget *mSTM;
- uint32_t mDeviceFlag;
+ uint32_t DeviceFlag;
private:
AMDGPUDeviceInfo::ExecutionMode
getExecutionMode(AMDGPUDeviceInfo::Caps Caps) const;
}; // AMDGPUDevice
} // namespace llvm
-#endif // _AMDILDEVICEIMPL_H_
+#endif // AMDILDEVICEIMPL_H
diff --git a/lib/Target/AMDGPU/AMDILDeviceInfo.h b/lib/Target/AMDGPU/AMDILDeviceInfo.h
index 4fa021e359..0eb3989776 100644
--- a/lib/Target/AMDGPU/AMDILDeviceInfo.h
+++ b/lib/Target/AMDGPU/AMDILDeviceInfo.h
@@ -6,18 +6,16 @@
// License. See LICENSE.TXT for details.
//
//==-----------------------------------------------------------------------===//
-#ifndef _AMDILDEVICEINFO_H_
-#define _AMDILDEVICEINFO_H_
+#ifndef AMDILDEVICEINFO_H
+#define AMDILDEVICEINFO_H
#include <string>
-namespace llvm
-{
+namespace llvm {
class AMDGPUDevice;
class AMDGPUSubtarget;
- namespace AMDGPUDeviceInfo
- {
+ namespace AMDGPUDeviceInfo {
// Each Capabilities can be executed using a hardware instruction,
// emulated with a sequence of software instructions, or not
// supported at all.
@@ -87,4 +85,4 @@ namespace llvm
bool is64bit = false, bool is64on32bit = false);
} // namespace AMDILDeviceInfo
} // namespace llvm
-#endif // _AMDILDEVICEINFO_H_
+#endif // AMDILDEVICEINFO_H
diff --git a/lib/Target/AMDGPU/AMDILDevices.h b/lib/Target/AMDGPU/AMDILDevices.h
index cfcc3304b4..befdf8f07d 100644
--- a/lib/Target/AMDGPU/AMDILDevices.h
+++ b/lib/Target/AMDGPU/AMDILDevices.h
@@ -6,8 +6,8 @@
// License. See LICENSE.TXT for details.
//
//==-----------------------------------------------------------------------===//
-#ifndef __AMDIL_DEVICES_H_
-#define __AMDIL_DEVICES_H_
+#ifndef AMDIL_DEVICES_H
+#define AMDIL_DEVICES_H
// Include all of the device specific header files
// This file is for Internal use only!
#include "AMDIL7XXDevice.h"
@@ -16,4 +16,4 @@
#include "AMDILNIDevice.h"
#include "AMDILSIDevice.h"
-#endif // _AMDIL_DEVICES_H_
+#endif // AMDIL_DEVICES_H
diff --git a/lib/Target/AMDGPU/AMDILEvergreenDevice.cpp b/lib/Target/AMDGPU/AMDILEvergreenDevice.cpp
index 28e6e84362..052238ce45 100644
--- a/lib/Target/AMDGPU/AMDILEvergreenDevice.cpp
+++ b/lib/Target/AMDGPU/AMDILEvergreenDevice.cpp
@@ -15,13 +15,13 @@ AMDGPUEvergreenDevice::AMDGPUEvergreenDevice(AMDGPUSubtarget *ST)
setCaps();
std::string name = ST->getDeviceName();
if (name == "cedar") {
- mDeviceFlag = OCL_DEVICE_CEDAR;
+ DeviceFlag = OCL_DEVICE_CEDAR;
} else if (name == "redwood") {
- mDeviceFlag = OCL_DEVICE_REDWOOD;
+ DeviceFlag = OCL_DEVICE_REDWOOD;
} else if (name == "cypress") {
- mDeviceFlag = OCL_DEVICE_CYPRESS;
+ DeviceFlag = OCL_DEVICE_CYPRESS;
} else {
- mDeviceFlag = OCL_DEVICE_JUNIPER;
+ DeviceFlag = OCL_DEVICE_JUNIPER;
}
}
diff --git a/lib/Target/AMDGPU/AMDILEvergreenDevice.h b/lib/Target/AMDGPU/AMDILEvergreenDevice.h
index 5def6955e1..d1a42ef122 100644
--- a/lib/Target/AMDGPU/AMDILEvergreenDevice.h
+++ b/lib/Target/AMDGPU/AMDILEvergreenDevice.h
@@ -14,8 +14,8 @@
// implement in order to correctly answer queries on the capabilities of the
// specific hardware.
//===----------------------------------------------------------------------===//
-#ifndef _AMDILEVERGREENDEVICE_H_
-#define _AMDILEVERGREENDEVICE_H_
+#ifndef AMDILEVERGREENDEVICE_H
+#define AMDILEVERGREENDEVICE_H
#include "AMDILDevice.h"
#include "AMDGPUSubtarget.h"
@@ -84,4 +84,4 @@ private:
}; // AMDGPURedwoodDevice
} // namespace llvm
-#endif // _AMDILEVERGREENDEVICE_H_
+#endif // AMDILEVERGREENDEVICE_H
diff --git a/lib/Target/AMDGPU/AMDILFrameLowering.h b/lib/Target/AMDGPU/AMDILFrameLowering.h
index 934ee46821..81c09a50d7 100644
--- a/lib/Target/AMDGPU/AMDILFrameLowering.h
+++ b/lib/Target/AMDGPU/AMDILFrameLowering.h
@@ -17,8 +17,8 @@
// Interface to describe a layout of a stack frame on a AMDIL target machine
//
//===----------------------------------------------------------------------===//
-#ifndef _AMDILFRAME_LOWERING_H_
-#define _AMDILFRAME_LOWERING_H_
+#ifndef AMDILFRAME_LOWERING_H
+#define AMDILFRAME_LOWERING_H
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/Target/TargetFrameLowering.h"
@@ -43,4 +43,4 @@ namespace llvm {
virtual bool hasFP(const MachineFunction &MF) const;
}; // class AMDGPUFrameLowering
} // namespace llvm
-#endif // _AMDILFRAME_LOWERING_H_
+#endif // AMDILFRAME_LOWERING_H
diff --git a/lib/Target/AMDGPU/AMDILISelDAGToDAG.cpp b/lib/Target/AMDGPU/AMDILISelDAGToDAG.cpp
index 2a80f1b25b..e7115bc778 100644
--- a/lib/Target/AMDGPU/AMDILISelDAGToDAG.cpp
+++ b/lib/Target/AMDGPU/AMDILISelDAGToDAG.cpp
@@ -157,110 +157,107 @@ SDNode *AMDGPUDAGToDAGISel::Select(SDNode *N) {
}
switch (Opc) {
default: break;
- case ISD::FrameIndex:
- {
- if (FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>(N)) {
- unsigned int FI = FIN->getIndex();
- EVT OpVT = N->getValueType(0);
- unsigned int NewOpc = AMDGPU::COPY;
- SDValue TFI = CurDAG->getTargetFrameIndex(FI, MVT::i32);
- return CurDAG->SelectNodeTo(N, NewOpc, OpVT, TFI);
- }
+ case ISD::FrameIndex: {
+ if (FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>(N)) {
+ unsigned int FI = FIN->getIndex();
+ EVT OpVT = N->getValueType(0);
+ unsigned int NewOpc = AMDGPU::COPY;
+ SDValue TFI = CurDAG->getTargetFrameIndex(FI, MVT::i32);
+ return CurDAG->SelectNodeTo(N, NewOpc, OpVT, TFI);
}
break;
+ }
case ISD::ConstantFP:
- case ISD::Constant:
- {
- const AMDGPUSubtarget &ST = TM.getSubtarget<AMDGPUSubtarget>();
- // XXX: Custom immediate lowering not implemented yet. Instead we use
- // pseudo instructions defined in SIInstructions.td
- if (ST.device()->getGeneration() > AMDGPUDeviceInfo::HD6XXX) {
- break;
+ case ISD::Constant: {
+ const AMDGPUSubtarget &ST = TM.getSubtarget<AMDGPUSubtarget>();
+ // XXX: Custom immediate lowering not implemented yet. Instead we use
+ // pseudo instructions defined in SIInstructions.td
+ if (ST.device()->getGeneration() > AMDGPUDeviceInfo::HD6XXX) {
+ break;
+ }
+ const R600InstrInfo *TII = static_cast<const R600InstrInfo*>(TM.getInstrInfo());
+
+ uint64_t ImmValue = 0;
+ unsigned ImmReg = AMDGPU::ALU_LITERAL_X;
+
+ if (N->getOpcode() == ISD::ConstantFP) {
+ // XXX: 64-bit Immediates not supported yet
+ assert(N->getValueType(0) != MVT::f64);
+
+ ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(N);
+ APFloat Value = C->getValueAPF();
+ float FloatValue = Value.convertToFloat();
+ if (FloatValue == 0.0) {
+ ImmReg = AMDGPU::ZERO;
+ } else if (FloatValue == 0.5) {
+ ImmReg = AMDGPU::HALF;
+ } else if (FloatValue == 1.0) {
+ ImmReg = AMDGPU::ONE;
+ } else {
+ ImmValue = Value.bitcastToAPInt().getZExtValue();
}
- const R600InstrInfo *TII = static_cast<const R600InstrInfo*>(TM.getInstrInfo());
-
- uint64_t ImmValue = 0;
- unsigned ImmReg = AMDGPU::ALU_LITERAL_X;
-
- if (N->getOpcode() == ISD::ConstantFP) {
- // XXX: 64-bit Immediates not supported yet
- assert(N->getValueType(0) != MVT::f64);
-
- ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(N);
- APFloat Value = C->getValueAPF();
- float FloatValue = Value.convertToFloat();
- if (FloatValue == 0.0) {
- ImmReg = AMDGPU::ZERO;
- } else if (FloatValue == 0.5) {
- ImmReg = AMDGPU::HALF;
- } else if (FloatValue == 1.0) {
- ImmReg = AMDGPU::ONE;
- } else {
- ImmValue = Value.bitcastToAPInt().getZExtValue();
- }
+ } else {
+ // XXX: 64-bit Immediates not supported yet
+ assert(N->getValueType(0) != MVT::i64);
+
+ ConstantSDNode *C = dyn_cast<ConstantSDNode>(N);
+ if (C->getZExtValue() == 0) {
+ ImmReg = AMDGPU::ZERO;
+ } else if (C->getZExtValue() == 1) {
+ ImmReg = AMDGPU::ONE_INT;
} else {
- // XXX: 64-bit Immediates not supported yet
- assert(N->getValueType(0) != MVT::i64);
-
- ConstantSDNode *C = dyn_cast<ConstantSDNode>(N);
- if (C->getZExtValue() == 0) {
- ImmReg = AMDGPU::ZERO;
- } else if (C->getZExtValue() == 1) {
- ImmReg = AMDGPU::ONE_INT;
- } else {
- ImmValue = C->getZExtValue();
- }
+ ImmValue = C->getZExtValue();
}
+ }
- for (SDNode::use_iterator Use = N->use_begin(), E = SDNode::use_end();
- Use != E; ++Use) {
- std::vector<SDValue> Ops;
- for (unsigned i = 0; i < Use->getNumOperands(); ++i) {
- Ops.push_back(Use->getOperand(i));
- }
+ for (SDNode::use_iterator Use = N->use_begin(), E = SDNode::use_end();
+ Use != E; ++Use) {
+ std::vector<SDValue> Ops;
+ for (unsigned i = 0; i < Use->getNumOperands(); ++i) {
+ Ops.push_back(Use->getOperand(i));
+ }
- if (!Use->isMachineOpcode()) {
- if (ImmReg == AMDGPU::ALU_LITERAL_X) {
- // We can only use literal constants (e.g. AMDGPU::ZERO,
- // AMDGPU::ONE, etc) in machine opcodes.
- continue;
- }
- } else {
- if (!TII->isALUInstr(Use->getMachineOpcode())) {
+ if (!Use->isMachineOpcode()) {
+ if (ImmReg == AMDGPU::ALU_LITERAL_X) {
+ // We can only use literal constants (e.g. AMDGPU::ZERO,
+ // AMDGPU::ONE, etc) in machine opcodes.
continue;
}
+ } else {
+ if (!TII->isALUInstr(Use->getMachineOpcode())) {
+ continue;
+ }
- int ImmIdx = TII->getOperandIdx(Use->getMachineOpcode(), R600Operands::IMM);
- assert(ImmIdx != -1);
-
- // subtract one from ImmIdx, because the DST operand is usually index
- // 0 for MachineInstrs, but we have no DST in the Ops vector.
- ImmIdx--;
+ int ImmIdx = TII->getOperandIdx(Use->getMachineOpcode(), R600Operands::IMM);
+ assert(ImmIdx != -1);
- // Check that we aren't already using an immediate.
- // XXX: It's possible for an instruction to have more than one
- // immediate operand, but this is not supported yet.
- if (ImmReg == AMDGPU::ALU_LITERAL_X) {
- ConstantSDNode *C = dyn_cast<ConstantSDNode>(Use->getOperand(ImmIdx));
- assert(C);
+ // subtract one from ImmIdx, because the DST operand is usually index
+ // 0 for MachineInstrs, but we have no DST in the Ops vector.
+ ImmIdx--;
- if (C->getZExtValue() != 0) {
- // This instruction is already using an immediate.
- continue;
- }
+ // Check that we aren't already using an immediate.
+ // XXX: It's possible for an instruction to have more than one
+ // immediate operand, but this is not supported yet.
+ if (ImmReg == AMDGPU::ALU_LITERAL_X) {
+ ConstantSDNode *C = dyn_cast<ConstantSDNode>(Use->getOperand(ImmIdx));
+ assert(C);
- // Set the immediate value
- Ops[ImmIdx] = CurDAG->getTargetConstant(ImmValue, MVT::i32);
+ if (C->getZExtValue() != 0) {
+ // This instruction is already using an immediate.
+ continue;
}
- }
- // Set the immediate register
- Ops[Use.getOperandNo()] = CurDAG->getRegister(ImmReg, MVT::i32);
- CurDAG->UpdateNodeOperands(*Use, Ops.data(), Use->getNumOperands());
+ // Set the immediate value
+ Ops[ImmIdx] = CurDAG->getTargetConstant(ImmValue, MVT::i32);
+ }
}
- break;
- }
+ // Set the immediate register
+ Ops[Use.getOperandNo()] = CurDAG->getRegister(ImmReg, MVT::i32);
+ CurDAG->UpdateNodeOperands(*Use, Ops.data(), Use->getNumOperands());
+ }
+ break;
+ }
}
return SelectCode(N);
}
@@ -388,8 +385,7 @@ bool AMDGPUDAGToDAGISel::isPrivateLoad(const LoadSDNode *N) {
&& !checkType(N->getSrcValue(), AMDGPUAS::REGION_ADDRESS)
&& !checkType(N->getSrcValue(), AMDGPUAS::CONSTANT_ADDRESS)
&& !checkType(N->getSrcValue(), AMDGPUAS::PARAM_D_ADDRESS)
- && !checkType(N->getSrcValue(), AMDGPUAS::PARAM_I_ADDRESS))
- {
+ && !checkType(N->getSrcValue(), AMDGPUAS::PARAM_I_ADDRESS)) {
return true;
}
return false;
diff --git a/lib/Target/AMDGPU/AMDILISelLowering.cpp b/lib/Target/AMDGPU/AMDILISelLowering.cpp
index 23bd92a658..e8cb703e75 100644
--- a/lib/Target/AMDGPU/AMDILISelLowering.cpp
+++ b/lib/Target/AMDGPU/AMDILISelLowering.cpp
@@ -44,8 +44,7 @@ using namespace llvm;
// TargetLowering Class Implementation Begins
//===----------------------------------------------------------------------===//
void AMDGPUTargetLowering::InitAMDILLowering() {
- int types[] =
- {
+ int types[] = {
(int)MVT::i8,
(int)MVT::i16,
(int)MVT::i32,
@@ -64,22 +63,19 @@ void AMDGPUTargetLowering::InitAMDILLowering() {
(int)MVT::v2i64
};
- int IntTypes[] =
- {
+ int IntTypes[] = {
(int)MVT::i8,
(int)MVT::i16,
(int)MVT::i32,
(int)MVT::i64
};
- int FloatTypes[] =
- {
+ int FloatTypes[] = {
(int)MVT::f32,
(int)MVT::f64
};
- int VectorTypes[] =
- {
+ int VectorTypes[] = {
(int)MVT::v2i8,
(int)MVT::v4i8,
(int)MVT::v2i16,
@@ -91,16 +87,16 @@ void AMDGPUTargetLowering::InitAMDILLowering() {
(int)MVT::v2f64,
(int)MVT::v2i64
};
- size_t numTypes = sizeof(types) / sizeof(*types);
- size_t numFloatTypes = sizeof(FloatTypes) / sizeof(*FloatTypes);
- size_t numIntTypes = sizeof(IntTypes) / sizeof(*IntTypes);
- size_t numVectorTypes = sizeof(VectorTypes) / sizeof(*VectorTypes);
+ size_t NumTypes = sizeof(types) / sizeof(*types);
+ size_t NumFloatTypes = sizeof(FloatTypes) / sizeof(*FloatTypes);
+ size_t NumIntTypes = sizeof(IntTypes) / sizeof(*IntTypes);
+ size_t NumVectorTypes = sizeof(VectorTypes) / sizeof(*VectorTypes);
const AMDGPUSubtarget &STM = getTargetMachine().getSubtarget<AMDGPUSubtarget>();
// These are the current register classes that are
// supported
- for (unsigned int x = 0; x < numTypes; ++x) {
+ for (unsigned int x = 0; x < NumTypes; ++x) {
MVT::SimpleValueType VT = (MVT::SimpleValueType)types[x];
//FIXME: SIGN_EXTEND_INREG is not meaningful for floating point types
@@ -121,7 +117,7 @@ void AMDGPUTargetLowering::InitAMDILLowering() {
setOperationAction(ISD::SDIV, VT, Custom);
}
}
- for (unsigned int x = 0; x < numFloatTypes; ++x) {
+ for (unsigned int x = 0; x < NumFloatTypes; ++x) {
MVT::SimpleValueType VT = (MVT::SimpleValueType)FloatTypes[x];
// IL does not have these operations for floating point types
@@ -136,7 +132,7 @@ void AMDGPUTargetLowering::InitAMDILLowering() {
setOperationAction(ISD::SETULE, VT, Expand);
}
- for (unsigned int x = 0; x < numIntTypes; ++x) {
+ for (unsigned int x = 0; x < NumIntTypes; ++x) {
MVT::SimpleValueType VT = (MVT::SimpleValueType)IntTypes[x];
// GPU also does not have divrem function for signed or unsigned
@@ -156,8 +152,7 @@ void AMDGPUTargetLowering::InitAMDILLowering() {
setOperationAction(ISD::CTLZ, VT, Expand);
}
- for ( unsigned int ii = 0; ii < numVectorTypes; ++ii )
- {
+ for (unsigned int ii = 0; ii < NumVectorTypes; ++ii) {
MVT::SimpleValueType VT = (MVT::SimpleValueType)VectorTypes[ii];
setOperationAction(ISD::VECTOR_SHUFFLE, VT, Expand);
@@ -229,10 +224,6 @@ void AMDGPUTargetLowering::InitAMDILLowering() {
maxStoresPerMemmove = 4096;
maxStoresPerMemset = 4096;
-#undef numTypes
-#undef numIntTypes
-#undef numVectorTypes
-#undef numFloatTypes
}
bool
diff --git a/lib/Target/AMDGPU/AMDILIntrinsicInfo.h b/lib/Target/AMDGPU/AMDILIntrinsicInfo.h
index 5d9e845af3..70c15d308a 100644
--- a/lib/Target/AMDGPU/AMDILIntrinsicInfo.h
+++ b/lib/Target/AMDGPU/AMDILIntrinsicInfo.h
@@ -10,8 +10,8 @@
// Interface for the AMDGPU Implementation of the Intrinsic Info class.
//
//===-----------------------------------------------------------------------===//
-#ifndef _AMDIL_INTRINSICS_H_
-#define _AMDIL_INTRINSICS_H_
+#ifndef AMDIL_INTRINSICS_H
+#define AMDIL_INTRINSICS_H
#include "llvm/Intrinsics.h"
#include "llvm/Target/TargetIntrinsicInfo.h"
@@ -43,5 +43,5 @@ namespace llvm {
}; // AMDGPUIntrinsicInfo
}
-#endif // _AMDIL_INTRINSICS_H_
+#endif // AMDIL_INTRINSICS_H
diff --git a/lib/Target/AMDGPU/AMDILNIDevice.cpp b/lib/Target/AMDGPU/AMDILNIDevice.cpp
index a903f440f9..6681e198c4 100644
--- a/lib/Target/AMDGPU/AMDILNIDevice.cpp
+++ b/lib/Target/AMDGPU/AMDILNIDevice.cpp
@@ -16,13 +16,13 @@ AMDGPUNIDevice::AMDGPUNIDevice(AMDGPUSubtarget *ST)
: AMDGPUEvergreenDevice(ST) {
std::string name = ST->getDeviceName();
if (name == "caicos") {
- mDeviceFlag = OCL_DEVICE_CAICOS;
+ DeviceFlag = OCL_DEVICE_CAICOS;
} else if (name == "turks") {
- mDeviceFlag = OCL_DEVICE_TURKS;
+ DeviceFlag = OCL_DEVICE_TURKS;
} else if (name == "cayman") {
- mDeviceFlag = OCL_DEVICE_CAYMAN;
+ DeviceFlag = OCL_DEVICE_CAYMAN;
} else {
- mDeviceFlag = OCL_DEVICE_BARTS;
+ DeviceFlag = OCL_DEVICE_BARTS;
}
}
AMDGPUNIDevice::~AMDGPUNIDevice() {
diff --git a/lib/Target/AMDGPU/AMDILNIDevice.h b/lib/Target/AMDGPU/AMDILNIDevice.h
index 387f7d1c3b..0579489c28 100644
--- a/lib/Target/AMDGPU/AMDILNIDevice.h
+++ b/lib/Target/AMDGPU/AMDILNIDevice.h
@@ -14,8 +14,8 @@
// implement in order to correctly answer queries on the capabilities of the
// specific hardware.
//===---------------------------------------------------------------------===//
-#ifndef _AMDILNIDEVICE_H_
-#define _AMDILNIDEVICE_H_
+#ifndef AMDILNIDEVICE_H
+#define AMDILNIDEVICE_H
#include "AMDILEvergreenDevice.h"
#include "AMDGPUSubtarget.h"
@@ -56,4 +56,4 @@ namespace llvm {
static const unsigned int MAX_LDS_SIZE_900 = AMDGPUDevice::MAX_LDS_SIZE_800;
} // namespace llvm
-#endif // _AMDILNIDEVICE_H_
+#endif // AMDILNIDEVICE_H
diff --git a/lib/Target/AMDGPU/AMDILPeepholeOptimizer.cpp b/lib/Target/AMDGPU/AMDILPeepholeOptimizer.cpp
index f4611f686f..6114ba3ba6 100644
--- a/lib/Target/AMDGPU/AMDILPeepholeOptimizer.cpp
+++ b/lib/Target/AMDGPU/AMDILPeepholeOptimizer.cpp
@@ -159,8 +159,7 @@ Function safeNestedForEach(InputIterator First, InputIterator Last,
namespace llvm {
FunctionPass *
- createAMDGPUPeepholeOpt(TargetMachine &tm)
- {
+ createAMDGPUPeepholeOpt(TargetMachine &tm) {
return new AMDGPUPeepholeOpt(tm);
}
} // llvm namespace
diff --git a/lib/Target/AMDGPU/AMDILRegisterInfo.td b/lib/Target/AMDGPU/AMDILRegisterInfo.td
index 42235ff37a..b9d033432e 100644
--- a/lib/Target/AMDGPU/AMDILRegisterInfo.td
+++ b/lib/Target/AMDGPU/AMDILRegisterInfo.td
@@ -85,24 +85,21 @@ def MEM : AMDILReg<999, "mem">, DwarfRegNum<[999]>;
def RA : AMDILReg<998, "r998">, DwarfRegNum<[998]>;
def FP : AMDILReg<997, "r997">, DwarfRegNum<[997]>;
def GPRI16 : RegisterClass<"AMDGPU", [i16], 16,
- (add (sequence "R%u", 1, 20), RA, SP, T1, T2, T3, T4, T5, SDP, R1010, R1011, R1001, R1002, R1003, R1004, R1005, R1006, R1007, R1008, MEM, R1012)>
-{
+ (add (sequence "R%u", 1, 20), RA, SP, T1, T2, T3, T4, T5, SDP, R1010, R1011, R1001, R1002, R1003, R1004, R1005, R1006, R1007, R1008, MEM, R1012)> {
let AltOrders = [(add (sequence "R%u", 1, 20))];
let AltOrderSelect = [{
return 1;
}];
}
def GPRI32 : RegisterClass<"AMDGPU", [i32], 32,
- (add (sequence "R%u", 1, 20), RA, SP, T1, T2, T3, T4, T5, SDP, R1010, R1011, R1001, R1002, R1003, R1004, R1005, R1006, R1007, R1008, MEM, R1012)>
-{
+ (add (sequence "R%u", 1, 20), RA, SP, T1, T2, T3, T4, T5, SDP, R1010, R1011, R1001, R1002, R1003, R1004, R1005, R1006, R1007, R1008, MEM, R1012)> {
let AltOrders = [(add (sequence "R%u", 1, 20))];
let AltOrderSelect = [{
return 1;
}];
}
def GPRF32 : RegisterClass<"AMDGPU", [f32], 32,
- (add (sequence "R%u", 1, 20), RA, SP, T1, T2, T3, T4, T5, SDP, R1010, R1011, R1001, R1002, R1003, R1004, R1005, R1006, R1007, R1008, MEM, R1012)>
-{
+ (add (sequence "R%u", 1, 20), RA, SP, T1, T2, T3, T4, T5, SDP, R1010, R1011, R1001, R1002, R1003, R1004, R1005, R1006, R1007, R1008, MEM, R1012)> {
let AltOrders = [(add (sequence "R%u", 1, 20))];
let AltOrderSelect = [{
return 1;
diff --git a/lib/Target/AMDGPU/AMDILSIDevice.h b/lib/Target/AMDGPU/AMDILSIDevice.h
index 6a684cb609..2272d85984 100644
--- a/lib/Target/AMDGPU/AMDILSIDevice.h
+++ b/lib/Target/AMDGPU/AMDILSIDevice.h
@@ -14,8 +14,8 @@
// implement in order to correctly answer queries on the capabilities of the
// specific hardware.
//===---------------------------------------------------------------------===//
-#ifndef _AMDILSIDEVICE_H_
-#define _AMDILSIDEVICE_H_
+#ifndef AMDILSIDEVICE_H
+#define AMDILSIDEVICE_H
#include "AMDILEvergreenDevice.h"
#include "AMDGPUSubtarget.h"
@@ -42,4 +42,4 @@ namespace llvm {
}; // AMDGPUSIDevice
} // namespace llvm
-#endif // _AMDILSIDEVICE_H_
+#endif // AMDILSIDEVICE_H
diff --git a/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.cpp b/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.cpp
index bbca34be3d..a0b7c321b7 100644
--- a/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.cpp
+++ b/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.cpp
@@ -10,8 +10,7 @@
#include "AMDGPUMCAsmInfo.h"
using namespace llvm;
-AMDGPUMCAsmInfo::AMDGPUMCAsmInfo(const Target &T, StringRef &TT) : MCAsmInfo()
-{
+AMDGPUMCAsmInfo::AMDGPUMCAsmInfo(const Target &T, StringRef &TT) : MCAsmInfo() {
HasSingleParameterDotFile = false;
WeakDefDirective = 0;
//===------------------------------------------------------------------===//
@@ -75,13 +74,11 @@ AMDGPUMCAsmInfo::AMDGPUMCAsmInfo(const Target &T, StringRef &TT) : MCAsmInfo()
}
const char*
-AMDGPUMCAsmInfo::getDataASDirective(unsigned int Size, unsigned int AS) const
-{
+AMDGPUMCAsmInfo::getDataASDirective(unsigned int Size, unsigned int AS) const {
return 0;
}
const MCSection*
-AMDGPUMCAsmInfo::getNonexecutableStackSection(MCContext &CTX) const
-{
+AMDGPUMCAsmInfo::getNonexecutableStackSection(MCContext &CTX) const {
return 0;
}
diff --git a/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.h b/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.h
index 0ca264beb0..d829f56b33 100644
--- a/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.h
+++ b/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.h
@@ -11,8 +11,8 @@
//
//===----------------------------------------------------------------------===//
-#ifndef AMDGPUMCASMINFO_H_
-#define AMDGPUMCASMINFO_H_
+#ifndef AMDGPUMCASMINFO_H
+#define AMDGPUMCASMINFO_H
#include "llvm/MC/MCAsmInfo.h"
namespace llvm {
@@ -27,4 +27,4 @@ namespace llvm {
const MCSection* getNonexecutableStackSection(MCContext &CTX) const;
};
} // namespace llvm
-#endif // AMDGPUMCASMINFO_H_
+#endif // AMDGPUMCASMINFO_H
diff --git a/lib/Target/AMDGPU/MCTargetDesc/R600MCCodeEmitter.cpp b/lib/Target/AMDGPU/MCTargetDesc/R600MCCodeEmitter.cpp
index 21b639b2fe..d61bc0b1d4 100644
--- a/lib/Target/AMDGPU/MCTargetDesc/R600MCCodeEmitter.cpp
+++ b/lib/Target/AMDGPU/MCTargetDesc/R600MCCodeEmitter.cpp
@@ -150,13 +150,12 @@ void R600MCCodeEmitter::EncodeInstruction(const MCInst &MI, raw_ostream &OS,
} else {
switch(MI.getOpcode()) {
case AMDGPU::RAT_WRITE_CACHELESS_32_eg:
- case AMDGPU::RAT_WRITE_CACHELESS_128_eg:
- {
- uint64_t inst = getBinaryCodeForInstr(MI, Fixups);
- EmitByte(INSTR_NATIVE, OS);
- Emit(inst, OS);
- break;
- }
+ case AMDGPU::RAT_WRITE_CACHELESS_128_eg: {
+ uint64_t inst = getBinaryCodeForInstr(MI, Fixups);
+ EmitByte(INSTR_NATIVE, OS);
+ Emit(inst, OS);
+ break;
+ }
case AMDGPU::CONSTANT_LOAD_eg:
case AMDGPU::VTX_READ_PARAM_i32_eg:
case AMDGPU::VTX_READ_PARAM_f32_eg:
@@ -164,24 +163,22 @@ void R600MCCodeEmitter::EncodeInstruction(const MCInst &MI, raw_ostream &OS,
case AMDGPU::VTX_READ_GLOBAL_i32_eg:
case AMDGPU::VTX_READ_GLOBAL_f32_eg:
case AMDGPU::VTX_READ_GLOBAL_v4i32_eg:
- case AMDGPU::VTX_READ_GLOBAL_v4f32_eg:
- {
- uint64_t InstWord01 = getBinaryCodeForInstr(MI, Fixups);
- uint32_t InstWord2 = MI.getOperand(2).getImm(); // Offset
-
- EmitByte(INSTR_VTX, OS);
- Emit(InstWord01, OS);
- Emit(InstWord2, OS);
- break;
- }
+ case AMDGPU::VTX_READ_GLOBAL_v4f32_eg: {
+ uint64_t InstWord01 = getBinaryCodeForInstr(MI, Fixups);
+ uint32_t InstWord2 = MI.getOperand(2).getImm(); // Offset
+
+ EmitByte(INSTR_VTX, OS);
+ Emit(InstWord01, OS);
+ Emit(InstWord2, OS);
+ break;
+ }
case AMDGPU::EG_Export:
- case AMDGPU::R600_Export:
- {
- uint64_t Inst = getBinaryCodeForInstr(MI, Fixups);
- EmitByte(INSTR_EXPORT, OS);
- Emit(Inst, OS);
- break;
- }
+ case AMDGPU::R600_Export: {
+ uint64_t Inst = getBinaryCodeForInstr(MI, Fixups);
+ EmitByte(INSTR_EXPORT, OS);
+ Emit(Inst, OS);
+ break;
+ }
default:
EmitALUInstr(MI, Fixups, OS);
diff --git a/lib/Target/AMDGPU/R600ExpandSpecialInstrs.cpp b/lib/Target/AMDGPU/R600ExpandSpecialInstrs.cpp
index df5c53afe9..dc73fa87e7 100644
--- a/lib/Target/AMDGPU/R600ExpandSpecialInstrs.cpp
+++ b/lib/Target/AMDGPU/R600ExpandSpecialInstrs.cpp
@@ -98,7 +98,7 @@ bool R600ExpandSpecialInstrsPass::ExpandInputPerspective(MachineInstr &MI) {
MI.getOperand(2).getImm());
- unsigned Sel;
+ unsigned Sel = AMDGPU::sel_x;
switch (i % 4) {
case 0:Sel = AMDGPU::sel_x;break;
case 1:Sel = AMDGPU::sel_y;break;
@@ -140,7 +140,7 @@ bool R600ExpandSpecialInstrsPass::ExpandInputConstant(MachineInstr &MI) {
unsigned ReadReg = AMDGPU::R600_ArrayBaseRegClass.getRegister(
MI.getOperand(1).getImm());
- unsigned Sel;
+ unsigned Sel = AMDGPU::sel_x;
switch (i % 4) {
case 0:Sel = AMDGPU::sel_x;break;
case 1:Sel = AMDGPU::sel_y;break;
@@ -220,7 +220,7 @@ bool R600ExpandSpecialInstrsPass::runOnMachineFunction(MachineFunction &MF) {
bool IsReduction = TII->isReductionOp(MI.getOpcode());
bool IsVector = TII->isVector(MI);
- bool IsCube = TII->isCubeOp(MI.getOpcode());
+ bool IsCube = TII->isCubeOp(MI.getOpcode());
if (!IsReduction && !IsVector && !IsCube) {
continue;
}
diff --git a/lib/Target/AMDGPU/R600ISelLowering.cpp b/lib/Target/AMDGPU/R600ISelLowering.cpp
index 18d3d1450d..d6c18b0750 100644
--- a/lib/Target/AMDGPU/R600ISelLowering.cpp
+++ b/lib/Target/AMDGPU/R600ISelLowering.cpp
@@ -80,53 +80,49 @@ MachineBasicBlock * R600TargetLowering::EmitInstrWithCustomInserter(
switch (MI->getOpcode()) {
default: return AMDGPUTargetLowering::EmitInstrWithCustomInserter(MI, BB);
case AMDGPU::SHADER_TYPE: break;
- case AMDGPU::CLAMP_R600:
- {
- MachineInstr *NewMI = TII->buildDefaultInstruction(*BB, I,
- AMDGPU::MOV,
- MI->getOperand(0).getReg(),
- MI->getOperand(1).getReg());
- TII->addFlag(NewMI, 0, MO_FLAG_CLAMP);
- break;
- }
- case AMDGPU::FABS_R600:
- {
- MachineInstr *NewMI = TII->buildDefaultInstruction(*BB, I,
+ case AMDGPU::CLAMP_R600: {
+ MachineInstr *NewMI = TII->buildDefaultInstruction(*BB, I,
+ AMDGPU::MOV,
+ MI->getOperand(0).getReg(),
+ MI->getOperand(1).getReg());
+ TII->addFlag(NewMI, 0, MO_FLAG_CLAMP);
+ break;
+ }
+
+ case AMDGPU::FABS_R600: {
+ MachineInstr *NewMI = TII->buildDefaultInstruction(*BB, I,
AMDGPU::MOV,
MI->getOperand(0).getReg(),
MI->getOperand(1).getReg());
- TII->addFlag(NewMI, 0, MO_FLAG_ABS);
- break;
- }
+ TII->addFlag(NewMI, 0, MO_FLAG_ABS);
+ break;
+ }
- case AMDGPU::FNEG_R600:
- {
- MachineInstr *NewMI = TII->buildDefaultInstruction(*BB, I,
+ case AMDGPU::FNEG_R600: {
+ MachineInstr *NewMI = TII->buildDefaultInstruction(*BB, I,
AMDGPU::MOV,
MI->getOperand(0).getReg(),
MI->getOperand(1).getReg());
- TII->addFlag(NewMI, 0, MO_FLAG_NEG);
+ TII->addFlag(NewMI, 0, MO_FLAG_NEG);
break;
- }
+ }
- case AMDGPU::R600_LOAD_CONST:
- {
- int64_t RegIndex = MI->getOperand(1).getImm();
- unsigned ConstantReg = AMDGPU::R600_CReg32RegClass.getRegister(RegIndex);
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::COPY))
- .addOperand(MI->getOperand(0))
- .addReg(ConstantReg);
- break;
- }
+ case AMDGPU::R600_LOAD_CONST: {
+ int64_t RegIndex = MI->getOperand(1).getImm();
+ unsigned ConstantReg = AMDGPU::R600_CReg32RegClass.getRegister(RegIndex);
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::COPY))
+ .addOperand(MI->getOperand(0))
+ .addReg(ConstantReg);
+ break;
+ }
- case AMDGPU::MASK_WRITE:
- {
- unsigned maskedRegister = MI->getOperand(0).getReg();
- assert(TargetRegisterInfo::isVirtualRegister(maskedRegister));
- MachineInstr * defInstr = MRI.getVRegDef(maskedRegister);
- TII->addFlag(defInstr, 0, MO_FLAG_MASK);
- break;
- }
+ case AMDGPU::MASK_WRITE: {
+ unsigned maskedRegister = MI->getOperand(0).getReg();
+ assert(TargetRegisterInfo::isVirtualRegister(maskedRegister));
+ MachineInstr * defInstr = MRI.getVRegDef(maskedRegister);
+ TII->addFlag(defInstr, 0, MO_FLAG_MASK);
+ break;
+ }
case AMDGPU::MOV_IMM_F32:
TII->buildMovImm(*BB, I, MI->getOperand(0).getReg(),
@@ -140,156 +136,154 @@ MachineBasicBlock * R600TargetLowering::EmitInstrWithCustomInserter(
case AMDGPU::RAT_WRITE_CACHELESS_32_eg:
- case AMDGPU::RAT_WRITE_CACHELESS_128_eg:
- {
- // Convert to DWORD address
- unsigned NewAddr = MRI.createVirtualRegister(
+ case AMDGPU::RAT_WRITE_CACHELESS_128_eg: {
+ // Convert to DWORD address
+ unsigned NewAddr = MRI.createVirtualRegister(
&AMDGPU::R600_TReg32_XRegClass);
- unsigned ShiftValue = MRI.createVirtualRegister(
+ unsigned ShiftValue = MRI.createVirtualRegister(
&AMDGPU::R600_TReg32RegClass);
- unsigned EOP = (llvm::next(I)->getOpcode() == AMDGPU::RETURN) ? 1 : 0;
-
- // XXX In theory, we should be able to pass ShiftValue directly to
- // the LSHR_eg instruction as an inline literal, but I tried doing it
- // this way and it didn't produce the correct results.
- TII->buildMovImm(*BB, I, ShiftValue, 2);
- TII->buildDefaultInstruction(*BB, I, AMDGPU::LSHR_eg, NewAddr,
- MI->getOperand(1).getReg(),
- ShiftValue);
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(MI->getOpcode()))
- .addOperand(MI->getOperand(0))
- .addReg(NewAddr)
- .addImm(EOP); // Set End of program bit
- break;
- }
+ unsigned EOP = (llvm::next(I)->getOpcode() == AMDGPU::RETURN) ? 1 : 0;
+
+ // XXX In theory, we should be able to pass ShiftValue directly to
+ // the LSHR_eg instruction as an inline literal, but I tried doing it
+ // this way and it didn't produce the correct results.
+ TII->buildMovImm(*BB, I, ShiftValue, 2);
+ TII->buildDefaultInstruction(*BB, I, AMDGPU::LSHR_eg, NewAddr,
+ MI->getOperand(1).getReg(),
+ ShiftValue);
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(MI->getOpcode()))
+ .addOperand(MI->getOperand(0))
+ .addReg(NewAddr)
+ .addImm(EOP); // Set End of program bit
+ break;
+ }
- case AMDGPU::RESERVE_REG:
- {
- R600MachineFunctionInfo * MFI = MF->getInfo<R600MachineFunctionInfo>();
- int64_t ReservedIndex = MI->getOperand(0).getImm();
- unsigned ReservedReg =
- AMDGPU::R600_TReg32RegClass.getRegister(ReservedIndex);
- MFI->ReservedRegs.push_back(ReservedReg);
- unsigned SuperReg =
+ case AMDGPU::RESERVE_REG: {
+ R600MachineFunctionInfo * MFI = MF->getInfo<R600MachineFunctionInfo>();
+ int64_t ReservedIndex = MI->getOperand(0).getImm();
+ unsigned ReservedReg =
+ AMDGPU::R600_TReg32RegClass.getRegister(ReservedIndex);
+ MFI->ReservedRegs.push_back(ReservedReg);
+ unsigned SuperReg =
AMDGPU::R600_Reg128RegClass.getRegister(ReservedIndex / 4);
- MFI->ReservedRegs.push_back(SuperReg);
- break;
- }
+ MFI->ReservedRegs.push_back(SuperReg);
+ break;
+ }
+
+ case AMDGPU::TXD: {
+ unsigned T0 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
+ unsigned T1 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
+
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_H), T0)
+ .addOperand(MI->getOperand(3))
+ .addOperand(MI->getOperand(4))
+ .addOperand(MI->getOperand(5))
+ .addOperand(MI->getOperand(6));
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_V), T1)
+ .addOperand(MI->getOperand(2))
+ .addOperand(MI->getOperand(4))
+ .addOperand(MI->getOperand(5))
+ .addOperand(MI->getOperand(6));
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SAMPLE_G))
+ .addOperand(MI->getOperand(0))
+ .addOperand(MI->getOperand(1))
+ .addOperand(MI->getOperand(4))
+ .addOperand(MI->getOperand(5))
+ .addOperand(MI->getOperand(6))
+ .addReg(T0, RegState::Implicit)
+ .addReg(T1, RegState::Implicit);
+ break;
+ }
+
+ case AMDGPU::TXD_SHADOW: {
+ unsigned T0 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
+ unsigned T1 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
+
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_H), T0)
+ .addOperand(MI->getOperand(3))
+ .addOperand(MI->getOperand(4))
+ .addOperand(MI->getOperand(5))
+ .addOperand(MI->getOperand(6));
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_V), T1)
+ .addOperand(MI->getOperand(2))
+ .addOperand(MI->getOperand(4))
+ .addOperand(MI->getOperand(5))
+ .addOperand(MI->getOperand(6));
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SAMPLE_C_G))
+ .addOperand(MI->getOperand(0))
+ .addOperand(MI->getOperand(1))
+ .addOperand(MI->getOperand(4))
+ .addOperand(MI->getOperand(5))
+ .addOperand(MI->getOperand(6))
+ .addReg(T0, RegState::Implicit)
+ .addReg(T1, RegState::Implicit);
+ break;
+ }
- case AMDGPU::TXD:
- {
- unsigned t0 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
- unsigned t1 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
-
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_H), t0)
- .addOperand(MI->getOperand(3))
- .addOperand(MI->getOperand(4))
- .addOperand(MI->getOperand(5))
- .addOperand(MI->getOperand(6));
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_V), t1)
- .addOperand(MI->getOperand(2))
- .addOperand(MI->getOperand(4))
- .addOperand(MI->getOperand(5))
- .addOperand(MI->getOperand(6));
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SAMPLE_G))
- .addOperand(MI->getOperand(0))
- .addOperand(MI->getOperand(1))
- .addOperand(MI->getOperand(4))
- .addOperand(MI->getOperand(5))
- .addOperand(MI->getOperand(6))
- .addReg(t0, RegState::Implicit)
- .addReg(t1, RegState::Implicit);
- break;
- }
- case AMDGPU::TXD_SHADOW:
- {
- unsigned t0 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
- unsigned t1 = MRI.createVirtualRegister(&AMDGPU::R600_Reg128RegClass);
-
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_H), t0)
- .addOperand(MI->getOperand(3))
- .addOperand(MI->getOperand(4))
- .addOperand(MI->getOperand(5))
- .addOperand(MI->getOperand(6));
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SET_GRADIENTS_V), t1)
- .addOperand(MI->getOperand(2))
- .addOperand(MI->getOperand(4))
- .addOperand(MI->getOperand(5))
- .addOperand(MI->getOperand(6));
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::TEX_SAMPLE_C_G))
- .addOperand(MI->getOperand(0))
- .addOperand(MI->getOperand(1))
- .addOperand(MI->getOperand(4))
- .addOperand(MI->getOperand(5))
- .addOperand(MI->getOperand(6))
- .addReg(t0, RegState::Implicit)
- .addReg(t1, RegState::Implicit);
- break;
- }
case AMDGPU::BRANCH:
BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::JUMP))
.addOperand(MI->getOperand(0))
.addReg(0);
break;
- case AMDGPU::BRANCH_COND_f32:
- {
- MachineInstr *NewMI =
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::PRED_X),
- AMDGPU::PREDICATE_BIT)
- .addOperand(MI->getOperand(1))
- .addImm(OPCODE_IS_NOT_ZERO)
- .addImm(0); // Flags
- TII->addFlag(NewMI, 0, MO_FLAG_PUSH);
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::JUMP))
- .addOperand(MI->getOperand(0))
- .addReg(AMDGPU::PREDICATE_BIT, RegState::Kill);
- break;
- }
- case AMDGPU::BRANCH_COND_i32:
- {
- MachineInstr *NewMI =
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::PRED_X),
+
+ case AMDGPU::BRANCH_COND_f32: {
+ MachineInstr *NewMI =
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::PRED_X),
AMDGPU::PREDICATE_BIT)
.addOperand(MI->getOperand(1))
- .addImm(OPCODE_IS_NOT_ZERO_INT)
+ .addImm(OPCODE_IS_NOT_ZERO)
.addImm(0); // Flags
- TII->addFlag(NewMI, 0, MO_FLAG_PUSH);
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::JUMP))
- .addOperand(MI->getOperand(0))
- .addReg(AMDGPU::PREDICATE_BIT, RegState::Kill);
+ TII->addFlag(NewMI, 0, MO_FLAG_PUSH);
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::JUMP))
+ .addOperand(MI->getOperand(0))
+ .addReg(AMDGPU::PREDICATE_BIT, RegState::Kill);
+ break;
+ }
+
+ case AMDGPU::BRANCH_COND_i32: {
+ MachineInstr *NewMI =
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::PRED_X),
+ AMDGPU::PREDICATE_BIT)
+ .addOperand(MI->getOperand(1))
+ .addImm(OPCODE_IS_NOT_ZERO_INT)
+ .addImm(0); // Flags
+ TII->addFlag(NewMI, 0, MO_FLAG_PUSH);
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(AMDGPU::JUMP))
+ .addOperand(MI->getOperand(0))
+ .addReg(AMDGPU::PREDICATE_BIT, RegState::Kill);
+ break;
+ }
+
+ case AMDGPU::input_perspective: {
+ R600MachineFunctionInfo *MFI = MF->getInfo<R600MachineFunctionInfo>();
+
+ // XXX Be more fine about register reservation
+ for (unsigned i = 0; i < 4; i ++) {
+ unsigned ReservedReg = AMDGPU::R600_TReg32RegClass.getRegister(i);
+ MFI->ReservedRegs.push_back(ReservedReg);
+ }
+
+ switch (MI->getOperand(1).getImm()) {
+ case 0:// Perspective
+ MFI->HasPerspectiveInterpolation = true;
break;
+ case 1:// Linear
+ MFI->HasLinearInterpolation = true;
+ break;
+ default:
+ assert(0 && "Unknow ij index");
}
- case AMDGPU::input_perspective:
- {
- R600MachineFunctionInfo *MFI = MF->getInfo<R600MachineFunctionInfo>();
-
- // XXX Be more fine about register reservation
- for (unsigned i = 0; i < 4; i ++) {
- unsigned ReservedReg = AMDGPU::R600_TReg32RegClass.getRegister(i);
- MFI->ReservedRegs.push_back(ReservedReg);
- }
- switch (MI->getOperand(1).getImm()) {
- case 0:// Perspective
- MFI->HasPerspectiveInterpolation = true;
- break;
- case 1:// Linear
- MFI->HasLinearInterpolation = true;
- break;
- default:
- assert(0 && "Unknow ij index");
- }
+ return BB;
+ }
- return BB;
- }
case AMDGPU::EG_Export:
- case AMDGPU::R600_Export:
- {
- bool EOP = (llvm::next(I)->getOpcode() == AMDGPU::RETURN)? 1 : 0;
- if (!EOP)
- return BB;
- unsigned CfInst = (MI->getOpcode() == AMDGPU::EG_Export)? 84 : 40;
- BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(MI->getOpcode()))
+ case AMDGPU::R600_Export: {
+ bool EOP = (llvm::next(I)->getOpcode() == AMDGPU::RETURN)? 1 : 0;
+ if (!EOP)
+ return BB;
+ unsigned CfInst = (MI->getOpcode() == AMDGPU::EG_Export)? 84 : 40;
+ BuildMI(*BB, I, BB->findDebugLoc(I), TII->get(MI->getOpcode()))
.addOperand(MI->getOperand(0))
.addOperand(MI->getOperand(1))
.addOperand(MI->getOperand(2))
@@ -299,8 +293,8 @@ MachineBasicBlock * R600TargetLowering::EmitInstrWithCustomInserter(
.addOperand(MI->getOperand(6))
.addImm(CfInst)
.addImm(1);
- break;
- }
+ break;
+ }
}
MI->eraseFromParent();
diff --git a/lib/Target/AMDGPU/R600ISelLowering.h b/lib/Target/AMDGPU/R600ISelLowering.h
index d1dfe9fe45..53e79884ec 100644
--- a/lib/Target/AMDGPU/R600ISelLowering.h
+++ b/lib/Target/AMDGPU/R600ISelLowering.h
@@ -20,8 +20,7 @@ namespace llvm {
class R600InstrInfo;
-class R600TargetLowering : public AMDGPUTargetLowering
-{
+class R600TargetLowering : public AMDGPUTargetLowering {
public:
R600TargetLowering(TargetMachine &TM);
virtual MachineBasicBlock * EmitInstrWithCustomInserter(MachineInstr *MI,
diff --git a/lib/Target/AMDGPU/R600InstrInfo.cpp b/lib/Target/AMDGPU/R600InstrInfo.cpp
index 814e0a2a7a..b9ade1e916 100644
--- a/lib/Target/AMDGPU/R600InstrInfo.cpp
+++ b/lib/Target/AMDGPU/R600InstrInfo.cpp
@@ -127,8 +127,7 @@ bool R600InstrInfo::isCubeOp(unsigned Opcode) const {
}
}
-bool R600InstrInfo::isALUInstr(unsigned Opcode) const
-{
+bool R600InstrInfo::isALUInstr(unsigned Opcode) const {
unsigned TargetFlags = get(Opcode).TSFlags;
return ((TargetFlags & R600_InstFlag::OP1) |
diff --git a/lib/Target/AMDGPU/R600Instructions.td b/lib/Target/AMDGPU/R600Instructions.td
index 35c0b6a814..39b49f6673 100644
--- a/lib/Target/AMDGPU/R600Instructions.td
+++ b/lib/Target/AMDGPU/R600Instructions.td
@@ -48,8 +48,7 @@ class InstR600 <bits<11> inst, dag outs, dag ins, string asm, list<dag> pattern,
}
class InstR600ISA <dag outs, dag ins, string asm, list<dag> pattern> :
- AMDGPUInst <outs, ins, asm, pattern>
-{
+ AMDGPUInst <outs, ins, asm, pattern> {
field bits<64> Inst;
let Namespace = "AMDGPU";
@@ -344,8 +343,7 @@ def TEX_SHADOW : PatLeaf<
class EG_CF_RAT <bits <8> cf_inst, bits <6> rat_inst, bits<4> rat_id, dag outs,
dag ins, string asm, list<dag> pattern> :
- InstR600ISA <outs, ins, asm, pattern>
-{
+ InstR600ISA <outs, ins, asm, pattern> {
bits<7> RW_GPR;
bits<7> INDEX_GPR;
@@ -410,8 +408,8 @@ def isCayman : Predicate<"Subtarget.device()"
"->getDeviceFlag() == OCL_DEVICE_CAYMAN">;
def isEGorCayman : Predicate<"Subtarget.device()"
"->getGeneration() == AMDGPUDeviceInfo::HD5XXX"
- "|| Subtarget.device()->getGeneration() =="
- "AMDGPUDeviceInfo::HD6XXX">;
+ "|| Subtarget.device()->getGeneration() =="
+ "AMDGPUDeviceInfo::HD6XXX">;
def isR600toCayman : Predicate<
"Subtarget.device()->getGeneration() <= AMDGPUDeviceInfo::HD6XXX">;
@@ -444,13 +442,11 @@ def input_constant : AMDGPUShaderInst <
-def INTERP_XY : R600_2OP <0xD6, "INTERP_XY", []>
-{
+def INTERP_XY : R600_2OP <0xD6, "INTERP_XY", []> {
let bank_swizzle = 5;
}
-def INTERP_ZW : R600_2OP <0xD7, "INTERP_ZW", []>
-{
+def INTERP_ZW : R600_2OP <0xD7, "INTERP_ZW", []> {
let bank_swizzle = 5;
}
@@ -499,8 +495,7 @@ class ExportInst : InstR600ISA<(
i32imm:$sw_x, i32imm:$sw_y, i32imm:$sw_z, i32imm:$sw_w, i32imm:$inst,
i32imm:$eop),
!strconcat("EXPORT", " $src"),
- []>
-{
+ []> {
bits<13> arraybase;
bits<2> type;
bits<7> src;
@@ -638,8 +633,8 @@ def SETGT_INT : R600_2OP <
>;
def SETGE_INT : R600_2OP <
- 0x3C, "SETGE_INT",
- [(set (i32 R600_Reg32:$dst),
+ 0x3C, "SETGE_INT",
+ [(set (i32 R600_Reg32:$dst),
(selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETGE))]
>;
@@ -667,7 +662,7 @@ def PRED_SETGE_INT : R600_2OP <0x44, "PRED_SETGE_INT", []>;
def PRED_SETNE_INT : R600_2OP <0x45, "PRED_SETNE_INT", []>;
def CNDE_INT : R600_3OP <
- 0x1C, "CNDE_INT",
+ 0x1C, "CNDE_INT",
[(set (i32 R600_Reg32:$dst),
(selectcc (i32 R600_Reg32:$src0), 0,
(i32 R600_Reg32:$src1), (i32 R600_Reg32:$src2),
@@ -675,7 +670,7 @@ def CNDE_INT : R600_3OP <
>;
def CNDGE_INT : R600_3OP <
- 0x1E, "CNDGE_INT",
+ 0x1E, "CNDGE_INT",
[(set (i32 R600_Reg32:$dst),
(selectcc (i32 R600_Reg32:$src0), 0,
(i32 R600_Reg32:$src1), (i32 R600_Reg32:$src2),
@@ -683,7 +678,7 @@ def CNDGE_INT : R600_3OP <
>;
def CNDGT_INT : R600_3OP <
- 0x1D, "CNDGT_INT",
+ 0x1D, "CNDGT_INT",
[(set (i32 R600_Reg32:$dst),
(selectcc (i32 R600_Reg32:$src0), 0,
(i32 R600_Reg32:$src1), (i32 R600_Reg32:$src2),
@@ -966,8 +961,7 @@ let Predicates = [isR600] in {
defm DIV_r600 : DIV_Common<RECIP_IEEE_r600>;
def TGSI_LIT_Z_r600 : TGSI_LIT_Z_Common<MUL_LIT_r600, LOG_CLAMPED_r600, EXP_IEEE_r600>;
- def R600_Export : ExportInst
- {
+ def R600_Export : ExportInst {
let Inst{52-49} = 1; // BURST_COUNT
let Inst{53} = eop;
let Inst{54} = 1; // VALID_PIXEL_MODE
@@ -1109,8 +1103,7 @@ let Predicates = [isEGorCayman] in {
def : Pat<(fp_to_uint R600_Reg32:$src0),
(FLT_TO_UINT_eg (TRUNC R600_Reg32:$src0))>;
- def EG_Export : ExportInst
- {
+ def EG_Export : ExportInst {
let Inst{51-48} = 1; // BURST_COUNT
let Inst{52} = 1; // VALID_PIXEL_MODE
let Inst{53} = eop;
@@ -1128,8 +1121,7 @@ let Predicates = [isEGorCayman] in {
let usesCustomInserter = 1 in {
class RAT_WRITE_CACHELESS_eg <dag ins, bits<4> comp_mask, string name> : EG_CF_RAT <
- 0x57, 0x2, 0, (outs), ins, !strconcat(name, " $rw_gpr, $index_gpr, $eop"), []>
-{
+ 0x57, 0x2, 0, (outs), ins, !strconcat(name, " $rw_gpr, $index_gpr, $eop"), []> {
let RIM = 0;
// XXX: Have a separate instruction for non-indexed writes.
let TYPE = 1;
diff --git a/lib/Target/AMDGPU/R600MachineFunctionInfo.cpp b/lib/Target/AMDGPU/R600MachineFunctionInfo.cpp
index c65b347b24..d710e3a4ef 100644
--- a/lib/Target/AMDGPU/R600MachineFunctionInfo.cpp
+++ b/lib/Target/AMDGPU/R600MachineFunctionInfo.cpp
@@ -14,8 +14,7 @@ using namespace llvm;
R600MachineFunctionInfo::R600MachineFunctionInfo(const MachineFunction &MF)
: MachineFunctionInfo(),
HasLinearInterpolation(false),
- HasPerspectiveInterpolation(false)
- {
+ HasPerspectiveInterpolation(false) {
memset(Outputs, 0, sizeof(Outputs));
}
diff --git a/lib/Target/AMDGPU/R600RegisterInfo.h b/lib/Target/AMDGPU/R600RegisterInfo.h
index e6ac2aa788..9188b937d5 100644
--- a/lib/Target/AMDGPU/R600RegisterInfo.h
+++ b/lib/Target/AMDGPU/R600RegisterInfo.h
@@ -22,8 +22,7 @@ namespace llvm {
class R600TargetMachine;
class TargetInstrInfo;
-struct R600RegisterInfo : public AMDGPURegisterInfo
-{
+struct R600RegisterInfo : public AMDGPURegisterInfo {
AMDGPUTargetMachine &TM;
const TargetInstrInfo &TII;
diff --git a/lib/Target/AMDGPU/SIFixSGPRLiveness.cpp b/lib/Target/AMDGPU/SIFixSGPRLiveness.cpp
index 56db55c563..de3a060d13 100644
--- a/lib/Target/AMDGPU/SIFixSGPRLiveness.cpp
+++ b/lib/Target/AMDGPU/SIFixSGPRLiveness.cpp
@@ -32,8 +32,7 @@ private:
MachineDominatorTree *MD;
MachinePostDominatorTree *MPD;
- bool isSGPR(const TargetRegisterClass *RegClass)
- {
+ bool isSGPR(const TargetRegisterClass *RegClass) {
return RegClass == &AMDGPU::SReg_1RegClass ||
RegClass == &AMDGPU::SReg_32RegClass ||
RegClass == &AMDGPU::SReg_64RegClass ||
diff --git a/lib/Target/AMDGPU/SIISelLowering.cpp b/lib/Target/AMDGPU/SIISelLowering.cpp
index d24db1ed81..889eeb4432 100644
--- a/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -179,7 +179,7 @@ void SITargetLowering::LowerSI_INTERP(MachineInstr *MI, MachineBasicBlock &BB,
.addOperand(iReg)
.addOperand(attr_chan)
.addOperand(attr)
- .addReg(M0);
+ .addReg(M0);
BuildMI(BB, I, BB.findDebugLoc(I), TII->get(AMDGPU::V_INTERP_P2_F32))
.addOperand(dst)
@@ -187,7 +187,7 @@ void SITargetLowering::LowerSI_INTERP(MachineInstr *MI, MachineBasicBlock &BB,
.addOperand(jReg)
.addOperand(attr_chan)
.addOperand(attr)
- .addReg(M0);
+ .addReg(M0);
MI->eraseFromParent();
}
@@ -208,7 +208,7 @@ void SITargetLowering::LowerSI_INTERP_CONST(MachineInstr *MI,
.addOperand(dst)
.addOperand(attr_chan)
.addOperand(attr)
- .addReg(M0);
+ .addReg(M0);
MI->eraseFromParent();
}
@@ -229,8 +229,8 @@ void SITargetLowering::LowerSI_V_CNDLT(MachineInstr *MI, MachineBasicBlock &BB,
unsigned VCC = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
BuildMI(BB, I, BB.findDebugLoc(I),
- TII->get(AMDGPU::V_CMP_GT_F32_e32),
- VCC)
+ TII->get(AMDGPU::V_CMP_GT_F32_e32),
+ VCC)
.addReg(AMDGPU::SREG_LIT_0)
.addOperand(MI->getOperand(1));
diff --git a/lib/Target/AMDGPU/SIISelLowering.h b/lib/Target/AMDGPU/SIISelLowering.h
index 4407bf0466..c82af86801 100644
--- a/lib/Target/AMDGPU/SIISelLowering.h
+++ b/lib/Target/AMDGPU/SIISelLowering.h
@@ -19,8 +19,7 @@
namespace llvm {
-class SITargetLowering : public AMDGPUTargetLowering
-{
+class SITargetLowering : public AMDGPUTargetLowering {
const SIInstrInfo * TII;
/// AppendS_WAITCNT - Memory reads and writes are syncronized using the
diff --git a/lib/Target/AMDGPU/SIInstrInfo.cpp b/lib/Target/AMDGPU/SIInstrInfo.cpp
index 60e7be4b26..3c773a0671 100644
--- a/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -47,7 +47,7 @@ SIInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
.addReg(SrcReg, getKillRegState(KillSrc));
} else if (AMDGPU::VReg_32RegClass.contains(DestReg)) {
assert(AMDGPU::VReg_32RegClass.contains(SrcReg) ||
- AMDGPU::SReg_32RegClass.contains(SrcReg));
+ AMDGPU::SReg_32RegClass.contains(SrcReg));
BuildMI(MBB, MI, DL, get(AMDGPU::V_MOV_B32_e32), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc));
} else {
diff --git a/lib/Target/AMDGPU/SIInstrInfo.td b/lib/Target/AMDGPU/SIInstrInfo.td
index ea8a33fb5a..873a451e99 100644
--- a/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/lib/Target/AMDGPU/SIInstrInfo.td
@@ -562,7 +562,7 @@ class MTBUF_Store_Helper <bits<3> op, string asm, RegisterClass regClass> : MTBU
}
multiclass SMRD_Helper <bits<5> op, string asm, RegisterClass dstClass,
- ValueType vt> {
+ ValueType vt> {
def _IMM : SMRD <
op,
(outs dstClass:$dst),
diff --git a/lib/Target/AMDGPU/SIInstructions.td b/lib/Target/AMDGPU/SIInstructions.td
index d6f71f6003..a2123a79f3 100644
--- a/lib/Target/AMDGPU/SIInstructions.td
+++ b/lib/Target/AMDGPU/SIInstructions.td
@@ -1058,10 +1058,10 @@ def LOAD_CONST : AMDGPUShaderInst <
let usesCustomInserter = 1 in {
def SI_V_CNDLT : InstSI <
- (outs VReg_32:$dst),
- (ins VReg_32:$src0, VReg_32:$src1, VReg_32:$src2),
- "SI_V_CNDLT $dst, $src0, $src1, $src2",
- [(set VReg_32:$dst, (int_AMDGPU_cndlt VReg_32:$src0, VReg_32:$src1, VReg_32:$src2))]
+ (outs VReg_32:$dst),
+ (ins VReg_32:$src0, VReg_32:$src1, VReg_32:$src2),
+ "SI_V_CNDLT $dst, $src0, $src1, $src2",
+ [(set VReg_32:$dst, (int_AMDGPU_cndlt VReg_32:$src0, VReg_32:$src1, VReg_32:$src2))]
>;
def SI_INTERP : InstSI <
@@ -1080,17 +1080,17 @@ def SI_INTERP_CONST : InstSI <
>;
def SI_KIL : InstSI <
- (outs),
- (ins VReg_32:$src),
- "SI_KIL $src",
- [(int_AMDGPU_kill VReg_32:$src)]
+ (outs),
+ (ins VReg_32:$src),
+ "SI_KIL $src",
+ [(int_AMDGPU_kill VReg_32:$src)]
>;
def SI_WQM : InstSI <
- (outs),
- (ins),
- "SI_WQM",
- [(int_SI_wqm)]
+ (outs),
+ (ins),
+ "SI_WQM",
+ [(int_SI_wqm)]
>;
} // end usesCustomInserter
diff --git a/lib/Target/AMDGPU/SILowerControlFlow.cpp b/lib/Target/AMDGPU/SILowerControlFlow.cpp
index f91e3ea643..5e7c3fa97e 100644
--- a/lib/Target/AMDGPU/SILowerControlFlow.cpp
+++ b/lib/Target/AMDGPU/SILowerControlFlow.cpp
@@ -146,8 +146,8 @@ bool SILowerControlFlowPass::runOnMachineFunction(MachineFunction &MF) {
.addReg(Reg);
freeReg(Reg);
- if (MF.getInfo<SIMachineFunctionInfo>()->ShaderType == ShaderType::PIXEL &&
- PredicateStack.empty()) {
+ if (MF.getInfo<SIMachineFunctionInfo>()->ShaderType == ShaderType::PIXEL &&
+ PredicateStack.empty()) {
// If the exec mask is non-zero, skip the next two instructions
BuildMI(MBB, I, MBB.findDebugLoc(I), TII->get(AMDGPU::S_CBRANCH_EXECNZ))
.addImm(3)
@@ -167,7 +167,7 @@ bool SILowerControlFlowPass::runOnMachineFunction(MachineFunction &MF) {
// ... and terminate wavefront
BuildMI(MBB, I, MBB.findDebugLoc(I), TII->get(AMDGPU::S_ENDPGM));
- }
+ }
MI.eraseFromParent();
break;
}
diff --git a/lib/Target/AMDGPU/SIMachineFunctionInfo.h b/lib/Target/AMDGPU/SIMachineFunctionInfo.h
index 68097b44c5..26902fa7a7 100644
--- a/lib/Target/AMDGPU/SIMachineFunctionInfo.h
+++ b/lib/Target/AMDGPU/SIMachineFunctionInfo.h
@@ -14,8 +14,8 @@
//===----------------------------------------------------------------------===//
-#ifndef _SIMACHINEFUNCTIONINFO_H_
-#define _SIMACHINEFUNCTIONINFO_H_
+#ifndef SIMACHINEFUNCTIONINFO_H_
+#define SIMACHINEFUNCTIONINFO_H_
#include "llvm/CodeGen/MachineFunction.h"
diff --git a/lib/Target/AMDGPU/SIRegisterInfo.h b/lib/Target/AMDGPU/SIRegisterInfo.h
index b571da9f6a..e70ab95dc2 100644
--- a/lib/Target/AMDGPU/SIRegisterInfo.h
+++ b/lib/Target/AMDGPU/SIRegisterInfo.h
@@ -22,8 +22,7 @@ namespace llvm {
class AMDGPUTargetMachine;
class TargetInstrInfo;
-struct SIRegisterInfo : public AMDGPURegisterInfo
-{
+struct SIRegisterInfo : public AMDGPURegisterInfo {
AMDGPUTargetMachine &TM;
const TargetInstrInfo &TII;