Starting with Oracle 11gR1 Oracle JVM includes Just-in-Time compiler. Its goal is to convert JVM bytecode into platform-specific native code based on the performance data gathered in run-time. It is stated in the documentation that JIT “enables much faster execution” – and this is, in general, true. There are ways to control the behavior of JIT, one way is described in the MOS Doc ID 1309858.1, and another one here.
Actually the way to control JIT (apart from completely disabling it with JAVA_JIT_ENABLED), is mentioned in the documentation – it is the new method, SET_NATIVE_COMPILER_OPTION in the DBMS_JAVA package. Unfortunately there’s no any information on what are possible incoming values to the procedure and their meaning. It looks like they are documented in the function DBMS_JAVA.NATIVE_COMPILER_OPTIONS. The way the function is implemented looks very odd to me, so I wrote a query to transform the returned table of VARCHAR2 into more readable/”query-able” form:
col option_name format a35
col opt_desc format a40 word_wrap
col opt_values format a30 word_wrap
col opt_format format a40 word_wrap
select option_name,
opt_desc,
opt_values,
opt_format
from
(select nvl(option_name, (lag(option_name ignore nulls) over (order by rn))) option_name,
start_of_group,
listagg(line, chr(10)) within group (order by rn) over (partition by gr) full_line
from
(select case when start_of_group = 1 then
trim(both '''' from substr(line, instr(line, '''')))
end option_name,
nvl(start_of_group, (lag(start_of_group ignore nulls) over (order by rn))) start_of_group,
sum(start_of_group) over (order by rn) gr,
line,
rn
from
(select case when option_line like 'Option %''' then 1
when option_line like 'Values:%' then 2
when option_line like 'Format:%' then 3
end start_of_group,
option_line line,
rownum rn
from table(dbms_java.native_compiler_options) t
where t.option_line is not null
)
)
)
pivot (max(full_line) for start_of_group in (1 as opt_desc, 2 as opt_values, 3 as opt_format))
order by option_name;
And here is query output from my Oracle 11.2.0.2 instance
OPTION_NAME OPT_DESC OPT_VALUES OPT_FORMAT
----------------------------------- ---------------------------------------- ------------------------------ ----------------------------------------
asyncPollLevel Option 'asyncPollLevel' Values: Format:
Level of polling used by generated code. SysEventsOnly - Poll on 'asyncPollLevel', '<value>'
Asycynchronous events (such as Ctrl-C System-events (such as GC)
interrupt) are polled for in only. This is default.
compiled code. This option specifies MethodEntry - Poll on each
how compiled-code should poll method-entry.
for these events. MethodEntryAndBranches - Poll
on each method-entry and
backwards-branch.
cpuFeatureFlagOverride Option 'cpuFeatureFlagOverride' Values: [int] Format:
override the value of the current 'cpuFeatureFlagOverride', '<value>'
cpu-feature flags
debugAssertLevel Option 'debugAssertLevel' Values: Format:
Level of debugging assertions used by None - No assertions. This is 'debugAssertLevel', '<value>'
the compiler. default.
Minimal - Minimal assertions.
All - All assertions.
debugPrettyPrintMIR Option 'debugPrettyPrintMIR' Values: { true, false } Format:
Pretty-print before/after MIR graphs 'debugPrettyPrintMIR', '<value>'
debugPrintAnnotatedAsm Option 'debugPrintAnnotatedAsm' Values: { true, false } Format:
Print annotated assembly listing for 'debugPrintAnnotatedAsm', '<value>'
generated code
debugPrintBytecodes Option 'debugPrintBytecodes' Values: { true, false } Format:
Print the input bytecodes 'debugPrintBytecodes', '<value>'
debugPrintClonedMIR Option 'debugPrintClonedMIR' Values: { true, false } Format:
Print Cloned MIR. 'debugPrintClonedMIR', '<value>'
debugPrintGCMap Option 'debugPrintGCMap' Values: { true, false } Format:
Print the GC Map 'debugPrintGCMap', '<value>'
debugPrintHeapSSAAfterRepairSSA Option 'debugPrintHeapSSAAfterRepairSSA' Values: { true, false } Format:
Print Heap-SSA after repair-SSA. 'debugPrintHeapSSAAfterRepairSSA',
'<value>'
debugPrintHeapSSAAfterToSSA Option 'debugPrintHeapSSAAfterToSSA' Values: { true, false } Format:
Print Heap-SSA after SSA conversion. 'debugPrintHeapSSAAfterToSSA', '<value>'
debugPrintHeapSSABeforeLSE Option 'debugPrintHeapSSABeforeLSE' Values: { true, false } Format:
Print HeapSSA before Load-Store 'debugPrintHeapSSABeforeLSE', '<value>'
Elimination.
debugPrintLIRAfterToLIR Option 'debugPrintLIRAfterToLIR' Values: { true, false } Format:
Print LIR after doing MIR-to-LIR 'debugPrintLIRAfterToLIR', '<value>'
conversion.
debugPrintLIRPseudoBlocks Option 'debugPrintLIRPseudoBlocks' Values: { true, false } Format:
Print LIR pseudo blocks when printing 'debugPrintLIRPseudoBlocks', '<value>'
LIR.
debugPrintLSEIPLattice Option 'debugPrintLSEIPLattice' Values: { true, false } Format:
Print Load-Store Elimination 'debugPrintLSEIPLattice', '<value>'
Index-propagation Lattice.
debugPrintMIRAfterABCD Option 'debugPrintMIRAfterABCD' Values: { true, false } Format:
Print MIR after Array Bounds-Check 'debugPrintMIRAfterABCD', '<value>'
removal.
debugPrintMIRAfterBranchSimplificat Option Values: { true, false } Format:
ion 'debugPrintMIRAfterBranchSimplification' 'debugPrintMIRAfterBranchSimplification'
Print MIR after Branch Simplification. , '<value>'
debugPrintMIRAfterCFGDiamond Option 'debugPrintMIRAfterCFGDiamond' Values: { true, false } Format:
Print MIR after CFG Diamond Analysis. 'debugPrintMIRAfterCFGDiamond',
'<value>'
debugPrintMIRAfterFromSSA Option 'debugPrintMIRAfterFromSSA' Values: { true, false } Format:
Print MIR after converting out of SSA. 'debugPrintMIRAfterFromSSA', '<value>'
debugPrintMIRAfterInlining Option 'debugPrintMIRAfterInlining' Values: { true, false } Format:
Print MIR after Method Inlining. 'debugPrintMIRAfterInlining', '<value>'
debugPrintMIRAfterLSE Option 'debugPrintMIRAfterLSE' Values: { true, false } Format:
Print MIR after Load-Store Elimination. 'debugPrintMIRAfterLSE', '<value>'
debugPrintMIRAfterLoopOpts Option 'debugPrintMIRAfterLoopOpts' Values: { true, false } Format:
Print MIR after Loop Optimizations. 'debugPrintMIRAfterLoopOpts', '<value>'
debugPrintMIRAfterOSDTransformation Option Values: { true, false } Format:
'debugPrintMIRAfterOSDTransformation' 'debugPrintMIRAfterOSDTransformation',
Print MIR after doing platform-specific '<value>'
transformation.
debugPrintMIRAfterRepairSSA Option 'debugPrintMIRAfterRepairSSA' Values: { true, false } Format:
Print MIR after repair-SSA. 'debugPrintMIRAfterRepairSSA', '<value>'
debugPrintMIRAfterToSSA Option 'debugPrintMIRAfterToSSA' Values: { true, false } Format:
Print MIR after SSA conversion. 'debugPrintMIRAfterToSSA', '<value>'
debugPrintMIRAfterTraceProfiles Option 'debugPrintMIRAfterTraceProfiles' Values: { true, false } Format:
Print MIR after doing TraceProfile 'debugPrintMIRAfterTraceProfiles',
analysis. '<value>'
debugPrintMIRAfterUnfactor Option 'debugPrintMIRAfterUnfactor' Values: { true, false } Format:
Print MIR after unfactorization. 'debugPrintMIRAfterUnfactor', '<value>'
debugPrintMIRBeforeABCD Option 'debugPrintMIRBeforeABCD' Values: { true, false } Format:
Print MIR before Array Bounds-Check 'debugPrintMIRBeforeABCD', '<value>'
removal.
debugPrintMIRBeforeBranchSimplifica Option Values: { true, false } Format:
tion 'debugPrintMIRBeforeBranchSimplification 'debugPrintMIRBeforeBranchSimplification
' ', '<value>'
Print MIR before Branch Simplification.
debugPrintMIRBeforeCFGDiamond Option 'debugPrintMIRBeforeCFGDiamond' Values: { true, false } Format:
Print MIR before CFG Diamond Analysis. 'debugPrintMIRBeforeCFGDiamond',
'<value>'
debugPrintMIRBeforeFromSSA Option 'debugPrintMIRBeforeFromSSA' Values: { true, false } Format:
Print MIR before converting out of SSA. 'debugPrintMIRBeforeFromSSA', '<value>'
debugPrintMIRBeforeGVN Option 'debugPrintMIRBeforeGVN' Values: { true, false } Format:
Print MIR before GlobalValueNumbering. 'debugPrintMIRBeforeGVN', '<value>'
debugPrintMIRBeforeInlining Option 'debugPrintMIRBeforeInlining' Values: { true, false } Format:
Print MIR before Method Inlining. 'debugPrintMIRBeforeInlining', '<value>'
debugPrintMIRBeforeLSE Option 'debugPrintMIRBeforeLSE' Values: { true, false } Format:
Print MIR before Load-Store Elimination. 'debugPrintMIRBeforeLSE', '<value>'
debugPrintMIRBeforeLoopOpts Option 'debugPrintMIRBeforeLoopOpts' Values: { true, false } Format:
Print MIR before Loop Optimizations. 'debugPrintMIRBeforeLoopOpts', '<value>'
debugPrintMIRBeforeOSDTransformatio Option Values: { true, false } Format:
n 'debugPrintMIRBeforeOSDTransformation' 'debugPrintMIRBeforeOSDTransformation',
Print MIR before doing platform-specific '<value>'
transformation.
debugPrintMIRBeforeRepairSSA Option 'debugPrintMIRBeforeRepairSSA' Values: { true, false } Format:
Print MIR before repair-SSA. 'debugPrintMIRBeforeRepairSSA',
'<value>'
debugPrintMIRBeforeToSSA Option 'debugPrintMIRBeforeToSSA' Values: { true, false } Format:
Print MIR before SSA conversion. 'debugPrintMIRBeforeToSSA', '<value>'
debugPrintMIRBeforeTraceProfiles Option Values: { true, false } Format:
'debugPrintMIRBeforeTraceProfiles' 'debugPrintMIRBeforeTraceProfiles',
Print MIR before doing TraceProfile '<value>'
analysis.
debugPrintMIRBeforeUnfactor Option 'debugPrintMIRBeforeUnfactor' Values: { true, false } Format:
Print MIR before unfactorization. 'debugPrintMIRBeforeUnfactor', '<value>'
debugTraceDataFlowSolver Option 'debugTraceDataFlowSolver' Values: { true, false } Format:
Trace the dataflow solver 'debugTraceDataFlowSolver', '<value>'
debugTraceLevel Option 'debugTraceLevel' Values: Format:
Level of debugging trace used by the None - No debugging. This is 'debugTraceLevel', '<value>'
compiler. default.
Minimal - Minimal debugging.
OptimizationsSummary - Minimal
debugging and summarize
optimizer decisions.
OptimizationsDetail - Minimal
debugging and detail optimizer
decisions.
Verbose - Verbose debugging.
All - All debugging.
exclude Option 'exclude' Format:
Exclude method from compilation. 'exclude', 'java.util.Hashtable'
'exclude',
'java.util.Hashtable.get(java.lang.Objec
t)'
'exclude',
'java.util.Hashtable.get(java.lang.Objec
t) return java.lang.Object'
'exclude',
'java/util/Hashtable.get(Ljava/lang/Obje
ct;)Ljava/lang/Object;
Java Virtual-Machine Specification style
method signatures,
javap-style method signatures and
call-spec style method signatures are
all
accepted.
If no method name/signature is given,
all methods in the indicated class
will be excluded
optimizerAnnotationConstraints Option 'optimizerAnnotationConstraints' Values: { true, false } Format:
Enable annotated class constraints. 'optimizerAnnotationConstraints',
'<value>'
optimizerBatchWKExceptions Option 'optimizerBatchWKExceptions' Values: { true, false } Format:
Enable batching of well-known (implicit) 'optimizerBatchWKExceptions', '<value>'
exception throws.
optimizerBoundsCheckRemoval Option 'optimizerBoundsCheckRemoval' Values: { true, false } Format:
Enable array bounds-check removal 'optimizerBoundsCheckRemoval', '<value>'
optimization.
optimizerBoundsCheckRemovalSubtract Option Values: { true, false } Format:
ionVertices 'optimizerBoundsCheckRemovalSubtractionV 'optimizerBoundsCheckRemovalSubtractionV
ertices' ertices', '<value>'
Enable subtraction vertices in array
bounds-check removal.
optimizerBranchElimination Option 'optimizerBranchElimination' Values: { true, false } Format:
Enable branch-elimination optimization. 'optimizerBranchElimination', '<value>'
optimizerBranchSimplification Option 'optimizerBranchSimplification' Values: { true, false } Format:
Enable branch-simplification 'optimizerBranchSimplification',
optimization. '<value>'
optimizerCombineDivMod Option 'optimizerCombineDivMod' Values: { true, false } Format:
Enable combination of div and mod 'optimizerCombineDivMod', '<value>'
operations.
optimizerCommonSubexpressionElimina Option Values: { true, false } Format:
tion 'optimizerCommonSubexpressionElimination 'optimizerCommonSubexpressionElimination
' ', '<value>'
Enable common-subexpression elimination
optimization.
optimizerConditionalMoves Option 'optimizerConditionalMoves' Values: { true, false } Format:
Enable analysis of CFG diamonds 'optimizerConditionalMoves', '<value>'
(conditional move optimization).
optimizerConstantFolding Option 'optimizerConstantFolding' Values: { true, false } Format:
Enable constant-folding optimization. 'optimizerConstantFolding', '<value>'
optimizerCopyPropagation Option 'optimizerCopyPropagation' Values: { true, false } Format:
Enable copy-propagation optimization. 'optimizerCopyPropagation', '<value>'
optimizerDeadCodeElimination Option 'optimizerDeadCodeElimination' Values: { true, false } Format:
Enable dead-code elimination 'optimizerDeadCodeElimination',
optimization. '<value>'
optimizerEagerlyCompileForInlining Option Values: { true, false } Format:
'optimizerEagerlyCompileForInlining' 'optimizerEagerlyCompileForInlining',
Enable eager inlining-driven '<value>'
compilation.
optimizerExceptionTraceHandling Option 'optimizerExceptionTraceHandling' Values: { true, false } Format:
Enable analysis of exception-trace 'optimizerExceptionTraceHandling',
usages to suppress unnecessary '<value>'
backtraces.
optimizerHnumConditionalMethodInlin Option Values: { true, false } Format:
ing 'optimizerHnumConditionalMethodInlining' 'optimizerHnumConditionalMethodInlining'
Enable hierarchy-number conditional , '<value>'
method-inlining.
optimizerInlineExceptionalMethods Option Values: { true, false } Format:
'optimizerInlineExceptionalMethods' 'optimizerInlineExceptionalMethods',
Enable inlining of methods that may '<value>'
potentially throw exceptions.
optimizerInlineExternalMethods Option 'optimizerInlineExternalMethods' Values: { true, false } Format:
Enable inlining of methods from other 'optimizerInlineExternalMethods',
classes. '<value>'
optimizerInlineInternedStrings Option 'optimizerInlineInternedStrings' Values: { true, false } Format:
Enable inlining of pointers to interned 'optimizerInlineInternedStrings',
strings. '<value>'
optimizerInlineNonLeafMethods Option 'optimizerInlineNonLeafMethods' Values: { true, false } Format:
Enable inlining of methods that are not 'optimizerInlineNonLeafMethods',
leaf methods. '<value>'
optimizerLoadStoreElimination Option 'optimizerLoadStoreElimination' Values: { true, false } Format:
Enable load-store elimination 'optimizerLoadStoreElimination',
optimization. '<value>'
optimizerLoopLICMOptimization Option 'optimizerLoopLICMOptimization' Values: { true, false } Format:
Enable loop LICM optimization. 'optimizerLoopLICMOptimization',
'<value>'
optimizerLoopPagedConversion Option 'optimizerLoopPagedConversion' Values: { true, false } Format:
Enable paged loop conversion. 'optimizerLoopPagedConversion',
'<value>'
optimizerLoopUnrollOptimization Option 'optimizerLoopUnrollOptimization' Values: { true, false } Format:
Enable loop Unroll optimization. 'optimizerLoopUnrollOptimization',
'<value>'
optimizerMethodInlining Option 'optimizerMethodInlining' Values: { true, false } Format:
Enable method-inlining. 'optimizerMethodInlining', '<value>'
optimizerOperationCombining Option 'optimizerOperationCombining' Values: { true, false } Format:
Enable combining of intermediate 'optimizerOperationCombining', '<value>'
representation operations
optimizerPartialRedundancyEliminati Option Values: { true, false } Format:
on 'optimizerPartialRedundancyElimination' 'optimizerPartialRedundancyElimination',
Enable partial redundancy elimination '<value>'
optimization.
optimizerSlowpathsOOL Option 'optimizerSlowpathsOOL' Values: { true, false } Format:
Enable out-of-line generation of 'optimizerSlowpathsOOL', '<value>'
slow-paths for implicit checks.
optimizerStrengthReduction Option 'optimizerStrengthReduction' Values: { true, false } Format:
Enable strength-reduction optimizations. 'optimizerStrengthReduction', '<value>'
optimizerThrowConversion Option 'optimizerThrowConversion' Values: { true, false } Format:
Enable conversion of local exception 'optimizerThrowConversion', '<value>'
throws to local jumps.
optimizerTypePropagation Option 'optimizerTypePropagation' Values: { true, false } Format:
Enable type-propagation. 'optimizerTypePropagation', '<value>'
optimizerUseAclintCache Option 'optimizerUseAclintCache' Values: { true, false } Format:
Enable use of aclint-cache. 'optimizerUseAclintCache', '<value>'
optimizerUseFastcalls Option 'optimizerUseFastcalls' Values: { true, false } Format:
Enable use of fastcalls when available. 'optimizerUseFastcalls', '<value>'
optimizerWellKnownMethods Option 'optimizerWellKnownMethods' Values: { true, false } Format:
Enable substitution of well-known method 'optimizerWellKnownMethods', '<value>'
implementations.
regallocCoalesceSpillLoadsStores Option Values: { true, false } Format:
'regallocCoalesceSpillLoadsStores' 'regallocCoalesceSpillLoadsStores',
Enable register allocator to '<value>'
aggressively coalesce spill loads and
stores.
regallocGenerateGCMaps Option 'regallocGenerateGCMaps' Values: { true, false } Format:
Enable register allocator to generate GC 'regallocGenerateGCMaps', '<value>'
maps for precise GC.
regallocPartitionRegisters Option 'regallocPartitionRegisters' Values: { true, false } Format:
Register allocator should partition 'regallocPartitionRegisters', '<value>'
registers according to whether or not
live-ranges span callsites.
scheduleInstructions Option 'scheduleInstructions' Values: { true, false } Format:
Enable instruction scheduler. 'scheduleInstructions', '<value>'
79 rows selected.
As you can see, some of the options look almost self-explanatory, like optimizerMethodInlining, some of them are cryptic. Well, that’s how it is. The most useful, I think, is exclude option which can be used to disable native compilation on a per-method basis. I haven’t tried it though, so, please use it carefully on your own risk.
