Nitro?

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Nitro?

Casey Ransberger-2
It would seem Safari is getting a new JS engine. Apple is touting "up to twice" the performance in Safari; that's almost certainly not coming out of a macro benchmark, though, so I strongly doubt that anyone with an app will see that much of a boost.

All the same, faster JS might be good for Lively on iOS, where perf is currently more pain than I can tolerate for more than a few minutes at a time.

This is supposed to be coming with iOS 4.3.
_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
Reply | Threaded
Open this post in threaded view
|

Re: Nitro?

Robert Krahn-3
Thanks for the info. I'm eager to try it out. In the not so far future Lively should run faster on Mobile Safari anyway.

Best,
Robert


On Mar 3, 2011, at 3:00 PM, Casey Ransberger wrote:

> It would seem Safari is getting a new JS engine. Apple is touting "up to twice" the performance in Safari; that's almost certainly not coming out of a macro benchmark, though, so I strongly doubt that anyone with an app will see that much of a boost.
>
> All the same, faster JS might be good for Lively on iOS, where perf is currently more pain than I can tolerate for more than a few minutes at a time.
>
> This is supposed to be coming with iOS 4.3.
> _______________________________________________
> lively-kernel mailing list
> [hidden email]
> http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel

_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
Reply | Threaded
Open this post in threaded view
|

Google Crankshaft (Re: Nitro?)

Antero Taivalsaari-2
Closely related to this topic:

Google's new Crankshaft compiler will be a standard
feature in Chrome 10 (the next major public release):

http://blog.chromium.org/2010/12/new-crankshaft-for-v8.html

Microsoft's Chakra VM in IE9 looks promising, too.

The race is still on...

-- Antero

> Thanks for the info. I'm eager to try it out. In the not so far future
> Lively should run faster on Mobile Safari anyway.
>
> Best,
> Robert
>
>
> On Mar 3, 2011, at 3:00 PM, Casey Ransberger wrote:
>
>> It would seem Safari is getting a new JS engine. Apple is touting "up to
>> twice" the performance in Safari; that's almost certainly not coming out
>> of a macro benchmark, though, so I strongly doubt that anyone with an
>> app will see that much of a boost.
>>
>> All the same, faster JS might be good for Lively on iOS, where perf is
>> currently more pain than I can tolerate for more than a few minutes at a
>> time.
>>
>> This is supposed to be coming with iOS 4.3.
>> _______________________________________________
>> lively-kernel mailing list
>> [hidden email]
>> http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
>
> _______________________________________________
> lively-kernel mailing list
> [hidden email]
> http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
>


_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
Reply | Threaded
Open this post in threaded view
|

Re: Nitro?

Dan Ingalls-4
In reply to this post by Robert Krahn-3
It's really true.

I ran my tiny benchmarks (http://weather-dimensions.com/Dan/JavaScriptBenchmark.html) before and after upgrading my iPad's OS.  The results...

Before: 30M ops/sec 1M sends/sec
After: 150M ops/sec 4M sends/sec

I found some erratic behavior running normal Lively pages -- we'll be tracking that down in the next couple of weeks.

I haven't had a chance to try a new iPad, but the extra processor should make a difference as well.

Think Live Web

        - Dan

>On Mar 3, 2011, at 3:00 PM, Casey Ransberger wrote:
>
>> It would seem Safari is getting a new JS engine. Apple is touting "up to twice" the performance in Safari; that's almost certainly not coming out of a macro benchmark, though, so I strongly doubt that anyone with an app will see that much of a boost.
>>
>> All the same, faster JS might be good for Lively on iOS, where perf is currently more pain than I can tolerate for more than a few minutes at a time.
>>
>> This is supposed to be coming with iOS 4.3.
>> _______________________________________________
>> lively-kernel mailing list
>> [hidden email]
>> http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
>
>_______________________________________________
>lively-kernel mailing list
>[hidden email]
>http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel

_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
Reply | Threaded
Open this post in threaded view
|

Re: Nitro?

alanone1
Hi Dan

Did you try the differential test that you did before? Or did Bert do it?

i.e.

Squeak vs JS-in-browser on a Mac

vs

the same on the iPad

I seem to recall that the differential was a factor of 12 slower on the iPad.

I visited Apple after this and complained to Bud Tribble.

Should I guess that the differential is now "only" a factor of 4?

Cheers,

Alan


From: Dan Ingalls <[hidden email]>
To: [hidden email]
Sent: Fri, March 11, 2011 3:45:14 PM
Subject: Re: [lively-kernel] Nitro?

It's really true.

I ran my tiny benchmarks (http://weather-dimensions.com/Dan/JavaScriptBenchmark.html) before and after upgrading my iPad's OS.  The results...

Before:    30M ops/sec    1M sends/sec
After:    150M ops/sec    4M sends/sec

I found some erratic behavior running normal Lively pages -- we'll be tracking that down in the next couple of weeks.

I haven't had a chance to try a new iPad, but the extra processor should make a difference as well.

Think Live Web

    - Dan

>On Mar 3, 2011, at 3:00 PM, Casey Ransberger wrote:
>
>> It would seem Safari is getting a new JS engine. Apple is touting "up to twice" the performance in Safari; that's almost certainly not coming out of a macro benchmark, though, so I strongly doubt that anyone with an app will see that much of a boost.
>>
>> All the same, faster JS might be good for Lively on iOS, where perf is currently more pain than I can tolerate for more than a few minutes at a time.
>>
>> This is supposed to be coming with iOS 4.3.
>> _______________________________________________
>> lively-kernel mailing list
>> [hidden email]
>> http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
>
>_______________________________________________
>lively-kernel mailing list
>[hidden email]
>http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel

_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel


_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
Reply | Threaded
Open this post in threaded view
|

Re: Nitro?

Dan Ingalls-4
Re: [lively-kernel] Nitro?
Hi Alan -

When I compare JS in Safari and Squeak native on my Mac (an old-ish 2.4 GHz core duo) I get:

Squeak native: 542M bytecodes/sec; 12.5M sends/sec
Cog native: 531M bytecodes/sec; 41M sends/sec
JS in Safari: 1000M ops/sec; 33M sends/sec
JS in Chrome: 4260M ops/sec; 49M sends/sec

My Safari is 5.0.4, from around January.  I believe it has the Nitro VM, but I'm not positive.
The Chrome is version 10.0.648.133

I designed the JS benchmark to compare JS and browser speeds, not to compare against Squeak.  We'd have to take some care with exactly how I did this before attributing a lot of significance.  I believe the sends are exactly comparable, but the 'ops' vs 'bytecodes' I'm not sure about.

It would appear that JS in the browsers is roughly the same speed as native Squeak at this point.

        - Dan
----------------------------------------------
Hi Dan

Did you try the differential test that you did before? Or did Bert do it?

i.e.

Squeak vs JS-in-browser on a Mac

vs

the same on the iPad

I seem to recall that the differential was a factor of 12 slower on the iPad.

I visited Apple after this and complained to Bud Tribble.

Should I guess that the differential is now "only" a factor of 4?

Cheers,

Alan


From: Dan Ingalls <[hidden email]>
To: [hidden email]
Sent: Fri, March 11, 2011 3:45:14 PM
Subject: Re: [lively-kernel] Nitro?

It's really true.

I ran my tiny benchmarks (http://weather-dimensions.com/Dan/JavaScriptBenchmark.html) before and after upgrading my iPad's OS.  The results...

Before:    30M ops/sec    1M sends/sec
After:    150M ops/sec    4M sends/sec

I found some erratic behavior running normal Lively pages -- we'll be tracking that down in the next couple of weeks.

I haven't had a chance to try a new iPad, but the extra processor should make a difference as well.

Think Live Web

    - Dan

>On Mar 3, 2011, at 3:00 PM, Casey Ransberger wrote:
>
>> It would seem Safari is getting a new JS engine. Apple is touting "up to twice" the performance in Safari; that's almost certainly not coming out of a macro benchmark, though, so I strongly doubt that anyone with an app will see that much of a boost.
>>
>> All the same, faster JS might be good for Lively on iOS, where perf is currently more pain than I can tolerate for more than a few minutes at a time.
>>
>> This is supposed to be coming with iOS 4.3.
>> _______________________________________________
>> lively-kernel mailing list
>> [hidden email]
>> http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel
>
>_______________________________________________
>lively-kernel mailing list
>[hidden email]
>http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel

_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel


_______________________________________________
lively-kernel mailing list
[hidden email]
http://lists.hpi.uni-potsdam.de/listinfo/lively-kernel