| PrintConfig | Boolean | false | Prints VM values (e.g. flags, constants, field offsets etc) exposed to JVMCI.
ArrayRegionEqualsConstantLimit = 4096 [community edition] [Integer]
Evaluates array region equality checks at compile time if the receiver
is a constant and the length of the array is less than this value.
AssemblyGCBarriers = true [community edition] [Boolean]
Use generated assembly for GC barriers if supported by the platform
BlindConstants = false [enterprise edition] [Boolean]
Blinds constants in code with a random key.
CompilationBailoutAsFailure = false [community edition] [Boolean]
Treats compilation bailouts as compilation failures.
CompilationFailureAction = Silent [community edition] [String]
Specifies the action to take when compilation fails.
The accepted values are:
Silent - Prints nothing to the console.
Print - Prints the stack trace to the console.
Diagnose* - Retries the compilation with extra diagnostics.
ExitVM - Same as Diagnose except that the VM process exits after retrying.
* If the value is "Diagnose", compilation is retried with extra diagnostics enabled including dumping.
Options specific to retry compilations can be modified using the DiagnoseOptions meta-option.
For example, to enable full debug dumping and logging during all retry compilations, use "-Djdk.graal.DiagnoseOptions=Dump=:5 Log=:5".
If the option value starts with a non-word character, that character is used as the separator between options instead of a space.
For example: "-Djdk.graal.DiagnoseOptions=@Log=Inlining@LogFile=/path/with space".
CompileGraalWithC1Only = true [community edition] [Boolean]
If in tiered mode, compiles Graal and JVMCI using optimized first-tier
code.
CompilerConfiguration = null [community edition] [String]
Names the compiler configuration to use. If omitted, uses the compiler
configuration with the greatest auto-selection priority. To see
available configurations, use the value 'help'.
ConditionalElimination = true [community edition] [Boolean]
Eliminates redundant conditional expressions and statements where
possible. This can improve performance because fewer logic instructions
have to be executed.
DeoptsToDisableOptimisticOptimization = 40 [community edition] [Integer]
Specifies the number of deoptimizations allowed per compilation unit
until optimistic optimizations are disabled.
DetectInvertedLoopsAsCounted = true [enterprise edition] [Boolean]
Applies counted loop optimization to tail-counted loops.
DiagnoseOptions = "Dump=:3" [community edition] [String]
Option values to use during a retry compilation triggered by
CompilationFailureAction=Diagnose or CompilationFailureAction=ExitVM.
If the value starts with a non-letter character, that character is used
as the separator between options instead of a space. For example:
"DiagnoseOptions=@Log=Inlining@LogFile=/path/with space."
DumpPath = "graal_dumps" [community edition] [String]
The directory where various Graal dump files are written.
DuplicationBudgetFactor = 0.25 [enterprise edition] [Double]
Specifies the percentage in node cost graph size for the duplication
budget (computed relative to the methods code size). The greater the
budget the more code duplication can be performed. This can improve
performance at the cost of additional code size.
EarlyExpandCheckCast = true [enterprise edition] [Boolean]
Tries to improve code generation by making checkcast operations better
optimizable early in the compilation pipeline.
EarlyGVN = true [community edition] [Boolean]
Performs early global value numbering on statements and expressions
directly after parsing. This can clean up the intermediate
representation and simplify later optimizations.
EarlyLICM = true [community edition] [Boolean]
Performs early loop-invariant code motion.
EmitStringSubstitutions = true [community edition] [Boolean]
Emits substitutions for String methods. This can improve performance
because the compiler can use optimized intrinsics for certain string
operations.
EnterpriseEarlyGVN = true [enterprise edition] [Boolean]
Performs an extended version of early global value numbering on a
graph. (Considers only graphs without floating guards.)
EnterprisePartialUnroll = true [enterprise edition] [Boolean]
Enables the advanced version of partial loop unrolling that considers
more loop shapes for unrolling.
EnterpriseRangeCheckElimination = true [enterprise edition] [Boolean]
Performs range check eliminations for Java long type range checks.
EnterpriseReadElimination = true [enterprise edition] [Boolean]
Runs more read eliminations early in the compilation pipeline.
FullUnroll = true [community edition] [Boolean]
Performs loop unrolling optimization.
Inline = true [community edition] [Boolean]
Performs inlining optimization. This can improve performance because
callees are specialized to the types and values of callers.
InlineDuringParsing = true [community edition] [Boolean]
Inlines trivial methods during bytecode parsing.
InlineGraalStubs = false [community edition] [Boolean]
Embeds all the emitted code for Graal-generated stubs.
InlinePartialIntrinsicExitDuringParsing = true [community edition][Boolean]
Inlines partial intrinsic exits during bytecode parsing when possible.
A partial intrinsic exit is a call within an intrinsic to the method
being intrinsified and denotes semantics of the original method that
the intrinsic does not support.
InlineVTableStubs = true [community edition] [Boolean]
Inlines the vtable stub for method dispatch during inlining.
Intrinsify = true [community edition] [Boolean]
Uses compiler intrinsifications.
LoopBoundOptimizationPhase = true [enterprise edition] [Boolean]
Attempts to improve counted loop detection by finding more precise loop
bounds.
LoopHeaderAlignment = 16 [community edition] [Integer]
Specifies the alignment in bytes for loop header blocks.
LoopInversion = true [enterprise edition] [Boolean]
Performs loop-inversion optimization.
LoopPeeling = true [community edition] [Boolean]
Performs loop peeling optimization.
LoopPredication = true [community edition] [Boolean]
Hoists array bounds checks out of simple loops. This is ignored if
SpeculativeGuardMovement is enabled.
LoopRotation = true [enterprise edition] [Boolean]
Enables loop rotation to let the compiler detect more loops as counted.
LoopUnswitch = true [community edition] [Boolean]
Performs loop unswitching optimization.
MaxCompilationProblemsPerAction = 2 [community edition] [Integer]
Specifies the maximum number of compilation failures to handle with the
action specified by CompilationFailureAction before changing to a less
verbose action. This does not apply to the ExitVM action..
MaxDuplicationFactor = 2.0 [community edition] [Double]
Specifies the maximum amount of extra effort to expend handling
irreducible loops. A value <= 1 disables support for irreducible loops.
MaximumEscapeAnalysisArrayLength = 128 [community edition] [Integer]
Specifies the maximum length of an array that will be escape analyzed.
MaximumInliningSize = 300 [community edition] [Integer]
Specifies the maximum graph size (measured in nodes) for which inlining
is explored for each call site.
MaximumRecursiveInlining = 5 [community edition] [Integer]
Specifies the maximum level of recursive inlining.
MinimalBulkZeroingSize = 2048 [community edition] [Integer]
If applicable, uses bulk zeroing instructions when the zeroing size in
bytes exceeds this threshold.
MinimumBlindedConstantSize = 4 [enterprise edition] [Integer]
Specifies the minimum size (in bytes) of constants to blind.
MoveGuardsUpwards = true [community edition] [Boolean]
Moves guard nodes to earlier places in the dominator tree if all
successors of a basic block share a common guard condition.
OptAssumptions = true [community edition] [Boolean]
Uses assumptions during compilation that may later be invalidated and
cause code to be deoptimized.
OptConditionalMoves = true [enterprise edition] [Boolean]
Optimizes simple if branches with conditional moves. This can improve
performance for patterns where branch prediction of a CPU does not work
(if branches have nearly equal probability).
OptConvertDeoptsToGuards = true [community edition] [Boolean]
Replaces deoptimization points with movable guards where possible. This
can help the optimizer to apply better code movement optimizations.
OptDeDuplication = true [enterprise edition] [Boolean]
Deduplicates statements and expressions before control flow merges if
they are equal. This can reduce code size.
OptDuplication = true [enterprise edition] [Boolean]
Performs statement level code duplication at control flow merges to
specialize code to branch values where possible.
OptExactArithmetic = true [enterprise edition] [Boolean]
Optimizes exact arithmetic where possible by rewriting it to non-exit
counterparts iff provably no overflow is possible.
OptFloatingReads = true [community edition] [Boolean]
Performs floating-read optimization. This enables memory read
operations to freely move in control-flow while respecting memory
(anti)-dependencies. This helps to reduce memory accesses and can
improve performance.
OptImplicitNullChecks = true [community edition] [Boolean]
Uses trapping null checks where possible to reduce explicit control
flow for null checks. This can improve performance because explicit
null checks do not have to be performed.
OptLockElimination = true [enterprise edition] [Boolean]
Tries to perform lock elimination where possible. Reduces the number of
monitorenter and monitorexit bytecodes if statically provable that they
are not needed.
OptMethodDuplication = false [enterprise edition] [Boolean]
Duplicates methods to form hot part and cold part areas.
OptReadElimination = true [community edition] [Boolean]
Tries to remove redundant memory accesses (for example, successive
reads of a non-volatile Java field).
OptWriteBarrierElimination = true [enterprise edition] [Boolean]
Eliminates redundant write barriers.
OptimisticAliasingAnalysis = true [enterprise edition] [Boolean]
Performs aliasing analysis on arrays to determine which memory does not
alias and enables more optimizations to be performed.
OptimizeLongJumps = false [community edition] [Boolean]
AMD64 only: Replace forward jumps (jmp, jcc) with equivalent but
smaller instructions if the actual jump displacement fits in one byte.
OptimizeLoopAccesses = true [enterprise edition] [Boolean]
Enables access node optimizations for loops. This can reduce the number
of memory operations executed in the body of a loop.
PartialEscapeAnalysis = true [community edition] [Boolean]
Performs partial escape analysis and scalar replacement optimization.
PartialRedundancyScheduling = true [enterprise edition] [Boolean]
Enables partial redundancy scheduling. This is a special form of code
scheduling that can revert the effects of partial redundancy
elimination (for example, global value numbering) by duplicating
expressions into branches. This can improve performance if partially
redundant expressions are only used in cold branches but the global
value numbered version not.
PartialUnroll = true [community edition] [Boolean]
Performs partial loop unrolling optimizations. This is a special form
of loop unrolling that splits a loop into a main and a post loop. The
main loop can then be unrolled by a fixed amount of iterations. The
post-loop performs any necessary fixup iterations. This can improve
performance because the loop control overhead is reduced in the
unrolled version.
ReassociateExpressions = true [community edition] [Boolean]
Reassociates loop invariants and constants.
RemoveLoopSafepoints = true [community edition] [Boolean]
Removes safepoints on counted loop ends.
RemoveNeverExecutedCode = true [community edition] [Boolean]
Uses deoptimization to prune branches of code in the generated code
that have never been executed by the interpreter.
ShowConfiguration = none [community edition] [String]
Writes the configuration of the selected compiler to the VM log.
SimulationBasedLoopPeeling = true [enterprise edition] [Boolean]
Uses the dominance-based duplication simulation (DBDS) algorithm to
simulate the impact of peeling on a loop.
SpectrePHTBarriers = None [community edition] [String]
Selects a strategy to mitigate speculative bounds check bypass (also
known as Spectre-PHT or Spectre V1).
The accepted values are:
None - No mitigations are used in JIT compiled code.
AllTargets - Speculative execution on all branch targets is
stopped using speculative execution barrier instructions.
This option is equivalent to setting SpeculativeExecutionBarriers to true.
GuardTargets - Branch targets relevant to Java memory safety are instrumented with barrier instructions.
This option has a lower performance impact than AllTargets.
NonDeoptGuardTargets - Same as GuardTargets, except that branches which deoptimize are not protected because they cannot be
executed repeatedly and are thus less likely to be successfully exploited in an attack.
Note that all modes except "None" also instrument branch target blocks containing UNSAFE memory accesses with barrier instructions.
SpectrePHTIndexMasking = false [community edition] [Boolean]
Masks indices to scope access to allocation size after bounds check.
SpeculativeGuardMovement = true [community edition] [Boolean]
Moves loop invariant guards (for example, array bounds checks) out of
loops.
SpeculativeStoreCheck = true [enterprise edition] [Boolean]
Speculates that arrays have exact type to optimize store checks. This
can improve performance because less store checks have to be performed.
(Dependent on loop invariant code motion.)
StringIndexOfConstantLimit = 4096 [community edition] [Integer]
Invocations of String.indexOf are evaluated at compile time if the
receiver is a constant and its length is less than this value.
StripMineCountedLoops = true [enterprise edition] [Boolean]
Performs strip-mining optimization on counted loops.
StripMineNonCountedLoops = true [enterprise edition] [Boolean]
Performs strip-mining optimization on non-counted loops.
SystemicCompilationFailureRate = 1 [community edition] [Integer]
Specifies the compilation failure rate that indicates a systemic
compilation problem (and a resulting warning). The value is made
absolute and clamped to produce P, a value between 0 and 100. Systemic
failure is detected if the percentage of failing compilations in a
sliding time window >= P. A negative value will cause the VM to exit
after issuing the warning. Set to 0 to disable systemic compilation
problem detection.
TrivialInliningSize = 10 [community edition] [Integer]
Specifies the size of a graph (counted in nodes) that is considered
trivial. Graphs with fewer than this number of nodes are therefore
always inlined.
TuneInlinerExploration = 0.0 [enterprise edition] [Double]
Increases or decreases the time spent exploring inlining opportunities
under the assumption that more time results in better peak performance
and less time reduces time to reach (a lower) peak performance. The
value of the option is clamped between -1 and 1 inclusively. A value
less than 0 reduces the exploration time and a value greater than 0
increases exploration time. Note that this option is only a heuristic
and should be tuned for any specific application.
TypicalGraphSize = 3250 [enterprise edition] [Integer]
Specifies the typical graph size at which inlining pressure must start
growing.
UseBranchesWithin32ByteBoundary = false [community edition] [Boolean]
Forces branch instructions to align with 32-bytes boundaries, to
mitigate the jcc erratum. See
https://www.intel.com/content/dam/support/us/en/documents/processors/mitigations-jump-conditional-code-erratum.pdf
for more details. If not set explicitly, the default value is
determined according to the CPU model..
UseTrappingNullChecks = true [community edition] [Boolean]
Uses traps for null checks instead of explicit null-checks. This can
improve performance because fewer explicit null checks have to be
performed.
VectorIntrinsics = true [enterprise edition] [Boolean]
Enables vectorized array copy intrinsics. This can improve performance
if the generated code uses vectorized intrinsics for array copy.
Vectorization = true [enterprise edition] [Boolean]
Enables vectorization. This is a global switch to enable/disable all
vectorization-related optimizations.
|
