V8 Performance Tests
Did a quick and dirty performance test comparing C, C++ and V8. This is not a valid benchmark in any way, it is only a test to give me a feeling about how "slow" V8 is compared to pure C and C++. The code also executed in the same loop and C/C++ tests was in a separate function so all 3 tests had exactly the same overhead in the C++ part. C : 1,4 Mill/sec C++: 0.8 Mill/sec V8 Run Only: 0,8 Mill/Sec V8 Compile+Run: 0,4 Mill/sec Other tests will show other differences - this is just an indication that V8 actually might compare with C/C++ performance wise + their memory handling look promissing. Threads ? They use globals and statics that will not thread that easy if you atteempt to thread V8 itself - I need to experiment and see. Jan
http://shootout.alioth.debian.org/u32/benchmark.php?test=all&lang=v8&lang2=g... ? From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Jan Berger Sent: Friday, July 02, 2010 5:45 PM To: freeswitch-dev@lists.freeswitch.org Subject: [Freeswitch-dev] V8 Performance Tests Did a quick and dirty performance test comparing C, C++ and V8. This is not a valid benchmark in any way, it is only a test to give me a feeling about how "slow" V8 is compared to pure C and C++. The code also executed in the same loop and C/C++ tests was in a separate function so all 3 tests had exactly the same overhead in the C++ part. C : 1,4 Mill/sec C++: 0.8 Mill/sec V8 Run Only: 0,8 Mill/Sec V8 Compile+Run: 0,4 Mill/sec Other tests will show other differences - this is just an indication that V8 actually might compare with C/C++ performance wise + their memory handling look promissing. Threads ? They use globals and statics that will not thread that easy if you atteempt to thread V8 itself - I need to experiment and see. Jan
These tests can be a bit missleading. Anyway - using the benchmark test on V8's own test below: Try running it on Chrome, and after that you try running it on Opera 10.60 The JavaScript in Opera 10.60 gives a higher score in my laptop. <http://v8.googlecode.com/svn/data/benchmarks/v5/run.html> http://v8.googlecode.com/svn/data/benchmarks/v5/run.html --- V8 is however really good from an implementation point of view. Using is as a script engine for an app is very straight forward unless you need threaded apps. V8 is a Multiprocessing by design, but it is not goof on multithreading. Threaded access to V8 is supported, but you need to lock V8 while access some parts of it (not all) - and it's a bit of work to get this part right. _____ From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Michael Giagnocavo Sent: 3. juli 2010 04:59 To: freeswitch-dev@lists.freeswitch.org Subject: Re: [Freeswitch-dev] V8 Performance Tests http://shootout.alioth.debian.org/u32/benchmark.php?test=all <http://shootout.alioth.debian.org/u32/benchmark.php?test=all&lang=v8&lang2= gcc> &lang=v8&lang2=gcc ? From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Jan Berger Sent: Friday, July 02, 2010 5:45 PM To: freeswitch-dev@lists.freeswitch.org Subject: [Freeswitch-dev] V8 Performance Tests Did a quick and dirty performance test comparing C, C++ and V8. This is not a valid benchmark in any way, it is only a test to give me a feeling about how "slow" V8 is compared to pure C and C++. The code also executed in the same loop and C/C++ tests was in a separate function so all 3 tests had exactly the same overhead in the C++ part. C : 1,4 Mill/sec C++: 0.8 Mill/sec V8 Run Only: 0,8 Mill/Sec V8 Compile+Run: 0,4 Mill/sec Other tests will show other differences - this is just an indication that V8 actually might compare with C/C++ performance wise + their memory handling look promissing. Threads ? They use globals and statics that will not thread that easy if you atteempt to thread V8 itself - I need to experiment and see. Jan
I don't doubt V8 is great. I'm just saying claiming that it can beat an optimizing C compiler on anything significant sounds a bit strange :) -Michael From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Jan Berger Sent: Sunday, July 04, 2010 2:26 AM To: freeswitch-dev@lists.freeswitch.org Subject: Re: [Freeswitch-dev] V8 Performance Tests These tests can be a bit missleading. Anyway - using the benchmark test on V8's own test below: Try running it on Chrome, and after that you try running it on Opera 10.60 The JavaScript in Opera 10.60 gives a higher score in my laptop. http://v8.googlecode.com/svn/data/benchmarks/v5/run.html --- V8 is however really good from an implementation point of view. Using is as a script engine for an app is very straight forward unless you need threaded apps. V8 is a Multiprocessing by design, but it is not goof on multithreading. Threaded access to V8 is supported, but you need to lock V8 while access some parts of it (not all) - and it's a bit of work to get this part right. ________________________________ From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Michael Giagnocavo Sent: 3. juli 2010 04:59 To: freeswitch-dev@lists.freeswitch.org Subject: Re: [Freeswitch-dev] V8 Performance Tests http://shootout.alioth.debian.org/u32/benchmark.php?test=all&lang=v8&lang2=g... ? From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Jan Berger Sent: Friday, July 02, 2010 5:45 PM To: freeswitch-dev@lists.freeswitch.org Subject: [Freeswitch-dev] V8 Performance Tests Did a quick and dirty performance test comparing C, C++ and V8. This is not a valid benchmark in any way, it is only a test to give me a feeling about how "slow" V8 is compared to pure C and C++. The code also executed in the same loop and C/C++ tests was in a separate function so all 3 tests had exactly the same overhead in the C++ part. C : 1,4 Mill/sec C++: 0.8 Mill/sec V8 Run Only: 0,8 Mill/Sec V8 Compile+Run: 0,4 Mill/sec Other tests will show other differences - this is just an indication that V8 actually might compare with C/C++ performance wise + their memory handling look promissing. Threads ? They use globals and statics that will not thread that easy if you atteempt to thread V8 itself - I need to experiment and see. Jan
My own tests below show that V8 is ca 30% of C if you include compilation, and 60% if you exclude it. I trust this test more as a brief handrule. . _____ From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Michael Giagnocavo Sent: 7. juli 2010 16:40 To: freeswitch-dev@lists.freeswitch.org Subject: Re: [Freeswitch-dev] V8 Performance Tests I don't doubt V8 is great. I'm just saying claiming that it can beat an optimizing C compiler on anything significant sounds a bit strange :-) -Michael From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Jan Berger Sent: Sunday, July 04, 2010 2:26 AM To: freeswitch-dev@lists.freeswitch.org Subject: Re: [Freeswitch-dev] V8 Performance Tests These tests can be a bit missleading. Anyway - using the benchmark test on V8's own test below: Try running it on Chrome, and after that you try running it on Opera 10.60 The JavaScript in Opera 10.60 gives a higher score in my laptop. <http://v8.googlecode.com/svn/data/benchmarks/v5/run.html> http://v8.googlecode.com/svn/data/benchmarks/v5/run.html --- V8 is however really good from an implementation point of view. Using is as a script engine for an app is very straight forward unless you need threaded apps. V8 is a Multiprocessing by design, but it is not goof on multithreading. Threaded access to V8 is supported, but you need to lock V8 while access some parts of it (not all) - and it's a bit of work to get this part right. _____ From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Michael Giagnocavo Sent: 3. juli 2010 04:59 To: freeswitch-dev@lists.freeswitch.org Subject: Re: [Freeswitch-dev] V8 Performance Tests http://shootout.alioth.debian.org/u32/benchmark.php?test=all <http://shootout.alioth.debian.org/u32/benchmark.php?test=all&lang=v8&lang2= gcc> &lang=v8&lang2=gcc ? From: freeswitch-dev-bounces@lists.freeswitch.org [mailto:freeswitch-dev-bounces@lists.freeswitch.org] On Behalf Of Jan Berger Sent: Friday, July 02, 2010 5:45 PM To: freeswitch-dev@lists.freeswitch.org Subject: [Freeswitch-dev] V8 Performance Tests Did a quick and dirty performance test comparing C, C++ and V8. This is not a valid benchmark in any way, it is only a test to give me a feeling about how "slow" V8 is compared to pure C and C++. The code also executed in the same loop and C/C++ tests was in a separate function so all 3 tests had exactly the same overhead in the C++ part. C : 1,4 Mill/sec C++: 0.8 Mill/sec V8 Run Only: 0,8 Mill/Sec V8 Compile+Run: 0,4 Mill/sec Other tests will show other differences - this is just an indication that V8 actually might compare with C/C++ performance wise + their memory handling look promissing. Threads ? They use globals and statics that will not thread that easy if you atteempt to thread V8 itself - I need to experiment and see. Jan
participants (2)
-
Jan Berger
-
Michael Giagnocavo