Hi Jakob, Yes, you're right. It was because of the debug build. Switching to the release version sped things up considerably. Thanks! That was a quick fix :)
Regards, Danny On Tuesday, March 24, 2015 at 11:09:27 AM UTC+2, Jakob Kummerow wrote: > > The profile for the "slow transaction" doesn't really say anything other > than that the C++ compiler didn't do any inlining (probably due to > oprofile, or was this a Debug build, or both?). > > > On Tue, Mar 24, 2015 at 9:50 AM, Toon Verwaest <verw...@chromium.org > <javascript:>> wrote: > >> If your transactional scripts contain function declarations that you >> instantiate, even if they are global, you'll inadvertantly cause >> polymorphism on the second run. That will slow down significantly. You can >> avoid this by only rerunning scripts that call/instantiate existing >> functions. Alternatively you can turn off the compilation cache; which will >> help with runtime (not load time though) if you reload all scripts. >> >> (Btw, you might want to turn off handle zapping as well...) >> >> Hth, >> Toon >> >> On Tue, Mar 24, 2015, 09:29 Danny Dorfman <wilder...@gmail.com >> <javascript:>> wrote: >> >>> Hello there, >>> >>> I have an issue with some code running JavaScript under V8, The code >>> runs in a *single* Isolate, that has *one* context. >>> The context is created, populated with some scripts, the scripts are >>> executed, and then the context is Exit()ed. The Isolate stays intact. >>> The above sequence repeats itself several times, with the *same* >>> scripts running time and again. This is what I call a series of >>> "transactions". >>> >>> What happens, is after the initial transaction is executed, all the >>> subsequent ones experience a drastic slowdown (sometimes 7 times slower). >>> I tried profiling the code with* oprofile*, and I see this picture: >>> >>> *Fast transaction:* >>> PU: Intel Ivy Bridge microarchitecture, speed 3401 MHz (estimated) >>> Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a >>> unit mask of 0x00 (No unit mask) count 100000 >>> samples % image name symbol name >>> 2444 1.8590 libv8.so >>> v8::internal::HandleScope::ZapRange(v8::internal::Object**, >>> v8::internal::Object**) >>> 1849 1.4064 libv8.so >>> v8::internal::HeapObject::map() const >>> 1537 1.1691 libv8.so v8::base::NoBarrier_Load(long >>> const volatile*) >>> 1417 1.0778 libv8.so >>> v8::internal::SmiTagging<8ul>::SmiToInt(v8::internal::Object const*) >>> 1286 0.9782 libv8.so >>> v8::internal::FixedArrayBase::length() const >>> 1273 0.9683 libpthread-2.13.so pthread_getspecific >>> 1250 0.9508 libv8.so >>> v8::internal::MapWord::MapWord(unsigned long) >>> 1235 0.9394 libv8.so >>> v8::internal::FixedArray::get(int) const >>> 1171 0.8907 libv8.so v8::base::Acquire_Load(long >>> const volatile*) >>> 1029 0.7827 libv8.so void >>> v8::base::CallOnce<void>(long*, v8::base::OneArgFunction<void*>::type, >>> void*) >>> 1021 0.7766 libv8.so >>> v8::internal::HeapObject::map_word() const >>> ..... >>> >>> *Slow transaction:* >>> CPU: Intel Ivy Bridge microarchitecture, speed 3401 MHz (estimated) >>> Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a >>> unit mask of 0x00 (No unit mask) count 100000 >>> samples % image name symbol name >>> 27474 8.4737 libv8.so >>> v8::internal::HeapObject::map() const >>> 17271 5.3268 libv8.so >>> v8::internal::MapWord::MapWord(unsigned long) >>> 15862 4.8923 libv8.so >>> v8::internal::Map::instance_type() >>> 14754 4.5505 libv8.so v8::base::NoBarrier_Load(long >>> const volatile*) >>> 13941 4.2998 libv8.so >>> v8::internal::HeapObject::map_word() const >>> 13573 4.1863 libv8.so >>> v8::internal::MapWord::FromRawValue(unsigned long) >>> 11299 3.4849 libv8.so >>> v8::internal::Internals::HasHeapObjectTag(v8::internal::Object const*) >>> 10365 3.1968 libv8.so >>> v8::internal::Object::IsInternalizedString() const >>> 8469 2.6121 libv8.so >>> v8::internal::HeapObject::cast(v8::internal::Object const*) >>> 8085 2.4936 libv8.so >>> v8::internal::Object::IsHeapObject() const >>> 6350 1.9585 libv8.so >>> v8::internal::Object::IsOddball() const >>> 6105 1.8829 libv8.so >>> v8::internal::Embedded::Embedded() >>> 5777 1.7818 libv8.so >>> v8::internal::MapWord::ToRawValue() >>> 5670 1.7488 libv8.so v8::internal::MapWord::ToMap() >>> ... >>> >>> Any idea what is happening here? Why is the same code running so slowly >>> the second (third, etc) time around? >>> >>> Regards, >>> Danny >>> >>> -- >>> -- >>> v8-users mailing list >>> v8-u...@googlegroups.com <javascript:> >>> http://groups.google.com/group/v8-users >>> --- >>> You received this message because you are subscribed to the Google >>> Groups "v8-users" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to v8-users+u...@googlegroups.com <javascript:>. >>> For more options, visit https://groups.google.com/d/optout. >>> >> -- >> -- >> v8-users mailing list >> v8-u...@googlegroups.com <javascript:> >> http://groups.google.com/group/v8-users >> --- >> You received this message because you are subscribed to the Google Groups >> "v8-users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to v8-users+u...@googlegroups.com <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > > -- -- v8-users mailing list v8-users@googlegroups.com http://groups.google.com/group/v8-users --- You received this message because you are subscribed to the Google Groups "v8-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to v8-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
