Comment #3 on issue 3860 by [email protected]: [v8 3.30.33.0] A segment
fault within deoptimization caused maybe by multi-thread
https://code.google.com/p/v8/issues/detail?id=3860
It happens only when turbofan is turned off on ARM, and not happens every
time.
And after futher research, i found that it may caused by "LAZY"
deoptimization.
If this "LAZY" was modified to "EARGR" or "SOFT", it works well.
void Deoptimizer::PatchCodeForDeoptimization(Isolate* isolate, Code* code) {
Address code_start_address = code->instruction_start();
// Invalidate the relocation information, as it will become invalid by the
// code patching below, and is not needed any more.
code->InvalidateRelocation();
...
for (int i = 0; i < deopt_data->DeoptCount(); i++) {
if (deopt_data->Pc(i)->value() == -1) continue;
Address call_address = code_start_address + deopt_data->Pc(i)->value();
Address deopt_entry = GetDeoptimizationEntry(isolate, i, LAZY);
//-----------LAZY
// We need calls to have a predictable size in the unoptimized code, but
// this is optimized code, so we don't have to have a predictable size.
int call_size_in_bytes =
MacroAssembler::CallSizeNotPredictableCodeSize(isolate,
deopt_entry,
RelocInfo::NONE32);
...
}
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
