org.graalvm.compiler.lir.aarch64

Class AArch64AtomicMove

java.lang.Object

org.graalvm.compiler.lir.aarch64.AArch64AtomicMove

public class AArch64AtomicMove
extends Object

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	AArch64AtomicMove.AtomicReadAndAddLSEOp Load (Read) and Add instruction.
static class	AArch64AtomicMove.AtomicReadAndAddOp Load (Read) and Add instruction.
static class	AArch64AtomicMove.AtomicReadAndWriteOp Load (Read) and Write instruction.
static class	AArch64AtomicMove.CompareAndSwapOp Compare and swap instruction.

Constructor Summary

Constructors

Constructor and Description
AArch64AtomicMove()

Method Summary

Modifier and Type	Method and Description
static AArch64LIRInstruction	createAtomicReadAndAdd(LIRGenerator gen, AArch64Kind kind, AllocatableValue result, AllocatableValue address, Value delta) Determines whether to use the atomic or load-store conditional implementation of atomic read&add based on the available hardware features.
static void	moveSPAndEmitCode(AArch64MacroAssembler masm, Register input, Consumer<Register> codeGen) This helper method moves sp to a scratch gp register before generating the code.

Methods inherited from class java.lang.Object

clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AArch64AtomicMove

public AArch64AtomicMove()

Method Detail

moveSPAndEmitCode

public static void moveSPAndEmitCode(AArch64MacroAssembler masm,
                                     Register input,
                                     Consumer<Register> codeGen)

This helper method moves sp to a scratch gp register before generating the code. This is needed when instructions within the codeGen cannot directly refer to the sp register. This can happen because both ZR and SP are encoded as '31'.

createAtomicReadAndAdd

public static AArch64LIRInstruction createAtomicReadAndAdd(LIRGenerator gen,
                                                           AArch64Kind kind,
                                                           AllocatableValue result,
                                                           AllocatableValue address,
                                                           Value delta)

Determines whether to use the atomic or load-store conditional implementation of atomic read&add based on the available hardware features. These two variants are split into two separate classes, as deltaValue is allowed to be a constant within the load-store conditional implementation.