DFG and FTL should support op_call_eval
https://bugs.webkit.org/show_bug.cgi?id=159786
Reviewed by Saam Barati.
Source/JavaScriptCore:
This adds support for op_call_eval in DFG and FTL by brute force:
- There is now a CallEval() node type, which compiles exactly the same way that we do in
baseline.
- We teach the DFG and bytecode liveness that the scope register and 'this' are read by
CallEval()/op_call_eval.
We can compile eval quite well, except that right now we cannot inline functions that use
eval. It would be nice to do that, but the payoff is probably smaller. "Don't inline users
of eval" may even be an OK inlining heuristic. Not inlining users of eval allows me to
reuse the baseline implementation, which is really great. Otherwise, I'd have to get rid
of things like the rogue reads of scope register and 'this'.
The goal here is to produce speed-ups for code that has functions that do both eval and
some computational stuff. Obviously, we're not producing any benefit for the eval itself.
But now the other stuff in a function that uses eval will get to participate in
optimization.
This is a huge speed-up on microbenchmarks.
* bytecode/BytecodeUseDef.h:
(JSC::computeUsesForBytecodeOffset):
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::printCallOp):
(JSC::CodeBlock::dumpBytecode):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::setLocal):
(JSC::DFG::ByteCodeParser::setArgument):
(JSC::DFG::ByteCodeParser::flush):
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGGraph.h:
(JSC::DFG::Graph::needsScopeRegister):
(JSC::DFG::Graph::needsFlushedThis):
* dfg/DFGHeapLocation.cpp:
(WTF::printInternal):
* dfg/DFGHeapLocation.h:
* dfg/DFGMayExit.cpp:
* dfg/DFGNode.h:
(JSC::DFG::Node::hasHeapPrediction):
* dfg/DFGNodeType.h:
* dfg/DFGOSRExitCompiler.cpp:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::emitCall):
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::emitCall):
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGStackLayoutPhase.cpp:
(JSC::DFG::StackLayoutPhase::run):
* dfg/DFGWatchpointCollectionPhase.cpp:
(JSC::DFG::WatchpointCollectionPhase::handle):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLCompile.cpp:
(JSC::FTL::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargs):
(JSC::FTL::DFG::LowerDFGToB3::compileCallEval):
(JSC::FTL::DFG::LowerDFGToB3::compileLoadVarargs):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::restoreCalleeSavesFromVMEntryFrameCalleeSavesBuffer):
(JSC::AssemblyHelpers::emitDumbVirtualCall):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::emitTypeOf):
* jit/JITCall.cpp:
(JSC::JIT::compileCallEvalSlowCase):
* jit/JITCall32_64.cpp:
(JSC::JIT::compileCallEvalSlowCase):
* jit/JITOperations.cpp:
* tests/stress/exit-then-eval.js: Added.
(foo):
* tests/stress/force-exit-then-eval-dfg.js: Added.
(foo):
* tests/stress/force-exit-then-eval.js: Added.
(foo):
LayoutTests:
* js/regress/eval-compute-expected.txt: Added.
* js/regress/eval-compute.html: Added.
* js/regress/eval-not-eval-compute-args-expected.txt: Added.
* js/regress/eval-not-eval-compute-args.html: Added.
* js/regress/eval-not-eval-compute-expected.txt: Added.
* js/regress/eval-not-eval-compute.html: Added.
* js/regress/script-tests/eval-compute.js: Added.
(foo):
* js/regress/script-tests/eval-not-eval-compute-args.js: Added.
(foo):
(i.result.foo):
* js/regress/script-tests/eval-not-eval-compute.js: Added.
(foo):
(i.result.foo):
git-svn-id: http://svn.webkit.org/repository/webkit/trunk@203364 268f45cc-cd09-0410-ab3c-d52691b4dbfc
43 files changed
