#include <instructions.hh>

Inheritance diagram for Gcn3ISA::Inst_SOPP__S_ENDPGM:

Public Member Functions
	Inst_SOPP__S_ENDPGM (InFmt_SOPP *)

	~Inst_SOPP__S_ENDPGM ()

int	getNumOperands () override

int	numDstRegOperands () override

int	numSrcRegOperands () override

int	getOperandSize (int opIdx) override

bool	isSrcOperand (int opIdx) override

bool	isDstOperand (int opIdx) override

void	execute (GPUDynInstPtr) override

Public Member Functions inherited from Gcn3ISA::Inst_SOPP
	Inst_SOPP (InFmt_SOPP *, const std::string &opcode)

	~Inst_SOPP ()

int	instSize () const override

void	generateDisassembly () override

bool	isScalarRegister (int opIdx) override

bool	isVectorRegister (int opIdx) override

int	getRegisterIndex (int opIdx, GPUDynInstPtr gpuDynInst) override

Public Member Functions inherited from Gcn3ISA::GCN3GPUStaticInst
	GCN3GPUStaticInst (const std::string &opcode)

	~GCN3GPUStaticInst ()

bool	isFlatScratchRegister (int opIdx) override

bool	isExecMaskRegister (int opIdx) override

int	coalescerTokenCount () const override
	Return the number of tokens needed by the coalescer. More...

ScalarRegU32	srcLiteral () const override

Public Member Functions inherited from GPUStaticInst
	GPUStaticInst (const std::string &opcode)

virtual	~GPUStaticInst ()

void	instAddr (int inst_addr)

int	instAddr () const

int	nextInstAddr () const

void	instNum (int num)

int	instNum ()

void	ipdInstNum (int num)

int	ipdInstNum () const

const std::string &	disassemble ()

int	numDstVecOperands ()

int	numSrcVecOperands ()

int	numDstVecDWORDs ()

int	numSrcVecDWORDs ()

int	numOpdDWORDs (int operandIdx)

bool	isALU () const

bool	isBranch () const

bool	isCondBranch () const

bool	isNop () const

bool	isReturn () const

bool	isEndOfKernel () const

bool	isKernelLaunch () const

bool	isSDWAInst () const

bool	isDPPInst () const

bool	isUnconditionalJump () const

bool	isSpecialOp () const

bool	isWaitcnt () const

bool	isBarrier () const

bool	isMemSync () const

bool	isMemRef () const

bool	isFlat () const

bool	isLoad () const

bool	isStore () const

bool	isAtomic () const

bool	isAtomicNoRet () const

bool	isAtomicRet () const

bool	isScalar () const

bool	readsSCC () const

bool	writesSCC () const

bool	readsVCC () const

bool	writesVCC () const

bool	readsEXEC () const

bool	writesEXEC () const

bool	readsMode () const

bool	writesMode () const

bool	ignoreExec () const

bool	isAtomicAnd () const

bool	isAtomicOr () const

bool	isAtomicXor () const

bool	isAtomicCAS () const

bool	isAtomicExch () const

bool	isAtomicAdd () const

bool	isAtomicSub () const

bool	isAtomicInc () const

bool	isAtomicDec () const

bool	isAtomicMax () const

bool	isAtomicMin () const

bool	isArgLoad () const

bool	isGlobalMem () const

bool	isLocalMem () const

bool	isArgSeg () const

bool	isGlobalSeg () const

bool	isGroupSeg () const

bool	isKernArgSeg () const

bool	isPrivateSeg () const

bool	isReadOnlySeg () const

bool	isSpillSeg () const

bool	isGloballyCoherent () const
	Coherence domain of a memory instruction. More...

bool	isSystemCoherent () const

bool	isF16 () const

bool	isF32 () const

bool	isF64 () const

bool	isFMA () const

bool	isMAC () const

bool	isMAD () const

virtual void	initiateAcc (GPUDynInstPtr gpuDynInst)

virtual void	completeAcc (GPUDynInstPtr gpuDynInst)

virtual uint32_t	getTargetPc ()

void	setFlag (Flags flag)

const std::string &	opcode () const

Additional Inherited Members
Public Attributes inherited from GPUStaticInst
Enums::StorageClassType	executed_as

Static Public Attributes inherited from GPUStaticInst
static uint64_t	dynamic_id_count

Protected Member Functions inherited from Gcn3ISA::GCN3GPUStaticInst
void	panicUnimplemented () const

Protected Attributes inherited from Gcn3ISA::Inst_SOPP
InFmt_SOPP	instData

Protected Attributes inherited from Gcn3ISA::GCN3GPUStaticInst
ScalarRegU32	_srcLiteral
	if the instruction has a src literal - an immediate value that is part of the instruction stream - we store that here More...

Protected Attributes inherited from GPUStaticInst
const std::string	_opcode

std::string	disassembly

int	_instNum

int	_instAddr

int	srcVecOperands

int	dstVecOperands

int	srcVecDWORDs

int	dstVecDWORDs

int	_ipdInstNum
	Identifier of the immediate post-dominator instruction. More...

std::bitset< Num_Flags >	_flags

Detailed Description

Definition at line 8398 of file instructions.hh.

Constructor & Destructor Documentation

◆ Inst_SOPP__S_ENDPGM()

Gcn3ISA::Inst_SOPP__S_ENDPGM::Inst_SOPP__S_ENDPGM ( InFmt_SOPP * iFmt )

Definition at line 3725 of file instructions.cc.

References GPUStaticInst::setFlag().

◆ ~Inst_SOPP__S_ENDPGM()

Gcn3ISA::Inst_SOPP__S_ENDPGM::~Inst_SOPP__S_ENDPGM ( )

Definition at line 3731 of file instructions.cc.

Member Function Documentation

◆ execute()

void Gcn3ISA::Inst_SOPP__S_ENDPGM::execute ( GPUDynInstPtr gpuDynInst )

overridevirtual

The parent WF of this instruction is exiting, therefore it should not participate in this barrier any longer. This prevents possible deadlock issues if WFs exit early.

If all WFs have finished, and hence the WG has finished, then we can free up the barrier belonging to the parent WG, but only if we actually used a barrier (i.e., more than one WF in the WG).

Last wavefront of the workgroup has executed return. If the workgroup is not the final one in the kernel, then simply retire it; however, if it is the final one (i.e., indicating the kernel end) then release operation is needed.

If a kernel end and release needed, inject a memory sync and retire the workgroup after receving all acks.

Implements GPUStaticInst.

Definition at line 3737 of file instructions.cc.

References ComputeUnit::activeWaves, Wavefront::barrierId(), ComputeUnit::completedWfs, ComputeUnit::completedWGs, Wavefront::computeUnit, ComputeUnit::cu_id, ComputeUnit::decMaxBarrierCnt(), LdsState::decreaseRefCounter(), Shader::dispatcher(), Wavefront::dispatchId, DPRINTF, Wavefront::dropFetch, ComputeUnit::fetchStage, FetchStage::fetchUnit(), FetchUnit::flushBuf(), RegisterManager::freeRegisters(), ComputeUnit::getLds(), Wavefront::getStatus(), Wavefront::hasBarrier(), ArmISA::i, Shader::impl_kern_end_rel, Wavefront::instructionBuffer, WFBarrier::InvalidID, GPUDispatcher::isReachingKernelEnd(), Wavefront::lastInstExec, ComputeUnit::maxBarrierCnt(), GPUDispatcher::notifyWgCompl(), panic_if, Wavefront::pendingFetch, Shader::prepareFlush(), Wavefront::rawDist, Wavefront::readsPerWrite, ComputeUnit::registerManager, Wavefront::releaseBarrier(), ComputeUnit::releaseBarrier(), Wavefront::S_BARRIER, Wavefront::S_RETURNING, Wavefront::S_STOPPED, Stats::DistBase< Derived, Stor >::sample(), GPUDispatcher::scheduleDispatch(), GPUStaticInst::setFlag(), Wavefront::setStatus(), ComputeUnit::shader, Wavefront::simdId, Wavefront::vecReads, Wavefront::wfDynId, Wavefront::wfSlotId, and Wavefront::wgId.