You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@pivot.apache.org by Piotr Kołaczkowski <pk...@gmail.com> on 2012/05/13 15:50:35 UTC
Synchronized resource cache
Hi,
I noticed that sometimes when we load components from BXML in
parallel, we get weird errors in resource cache. Making it
synchronized seems to fix it.
Patch attached.
Thanks,
Piotr
Re: Synchronized resource cache
Posted by Roger Whitcomb <Ro...@rbwhitcomb.com>.
So, Piotr, is what is in SVN now good (after Sandro's big patch)?
~Roger
Sent from my iPhone
On May 15, 2012, at 7:25 AM, Sandro Martini <sa...@gmail.com>
wrote:
> Hi Piotr,
>
>> I don't think this would help. If a thread iterates over a
>> collection while another thread is modifying it, you'd get
>> ConcurrentModificationException anyway.
> Just to be sure of this ... the code returns a new Iterator so we
> should be ok.
>
> public Iterator<URL> iterator() {
> return new Iterator<URL>() {
> private Iterator<URI> iterator =
> resourceCache.iterator();
> ...
> Some info on Iterators and multithread even here (just for reference):
> http://www.oodesign.com/iterator-pattern.html
>
>> Is this iterator ever used somewhere? If it is, we have to solve it
>> somehow differently. Nevertheless making just puts and gets
>> synchronized solved the issue at least for us - we don't encounter
>> any crashes while starting the app.
> Ok, so I think all should be good ...
>
>
>> The scenario that was causing this was the following:
>>
>> Thread one: load window 1 from BXML
>> Thread two: load window 2 from BXML
>> Thread three: load window 3 from BXML
>> Thread four: load window 4 from BXML
>> Join threads
>> In GUI thread - open all the loaded windows
> This could be a very interesting new Tutorial / Sample, if you have
> some time it would be great to prototype something.
> Or if someone of us has some time ... if possible I'll try in next
> days, but if you have something ready to use please tell me.
>
>> It is important for us to make it work (stable), because on a
>> multicore machine it improves loading windows by about 40% (not
>> perfect, but visibly faster).
> Yes, I understand, and it's a major problem in all modern software ...
> better usage of multicores ... and it's one of the reasons why I like
> so much Scala :-) .
>
>
>> BTW: Have you ever optimised pivot for faster startup times? The
>> first thing we noticed when we switched from Swing is that Pivot
>> was somehow slower to start up. This is worrying me - now it is not
>> a problem it the user waits for 1 second to load the app (I'm
>> talking about warm start up), but if in future the gui gets bigger...
> No, I don't think none of us has even done something like this ... an
> in-depth look with a Profiler would be welcome (mabye even starting
> with the NetBeans profiler, and/or VisualVM). An issue for 2.1 on this
> could be add.
>
>> The competition using JavaScript doesn't have that problem. :
>> ( Maybe there are some possibilities to limit the amount of
>> classloading or make some things more lazy?
> Yes, I know ... maybe for 2.1 we could really try to do something, but
> before someone of us should do some analysis with a profiler, to
> better understand if/where we can improve ...
>
> Bye,
> Sandro
>
Re: Synchronized resource cache
Posted by Sandro Martini <sa...@gmail.com>.
Hi Piotr,
> I don't think this would help. If a thread iterates over a collection while another thread is modifying it, you'd get ConcurrentModificationException anyway.
Just to be sure of this ... the code returns a new Iterator so we should be ok.
public Iterator<URL> iterator() {
return new Iterator<URL>() {
private Iterator<URI> iterator = resourceCache.iterator();
...
Some info on Iterators and multithread even here (just for reference):
http://www.oodesign.com/iterator-pattern.html
> Is this iterator ever used somewhere? If it is, we have to solve it somehow differently. Nevertheless making just puts and gets synchronized solved the issue at least for us - we don't encounter any crashes while starting the app.
Ok, so I think all should be good ...
> The scenario that was causing this was the following:
>
> Thread one: load window 1 from BXML
> Thread two: load window 2 from BXML
> Thread three: load window 3 from BXML
> Thread four: load window 4 from BXML
> Join threads
> In GUI thread - open all the loaded windows
This could be a very interesting new Tutorial / Sample, if you have
some time it would be great to prototype something.
Or if someone of us has some time ... if possible I'll try in next
days, but if you have something ready to use please tell me.
> It is important for us to make it work (stable), because on a multicore machine it improves loading windows by about 40% (not perfect, but visibly faster).
Yes, I understand, and it's a major problem in all modern software ...
better usage of multicores ... and it's one of the reasons why I like
so much Scala :-) .
> BTW: Have you ever optimised pivot for faster startup times? The first thing we noticed when we switched from Swing is that Pivot was somehow slower to start up. This is worrying me - now it is not a problem it the user waits for 1 second to load the app (I'm talking about warm start up), but if in future the gui gets bigger...
No, I don't think none of us has even done something like this ... an
in-depth look with a Profiler would be welcome (mabye even starting
with the NetBeans profiler, and/or VisualVM). An issue for 2.1 on this
could be add.
> The competition using JavaScript doesn't have that problem. :( Maybe there are some possibilities to limit the amount of classloading or make some things more lazy?
Yes, I know ... maybe for 2.1 we could really try to do something, but
before someone of us should do some analysis with a profiler, to
better understand if/where we can improve ...
Bye,
Sandro
Re: Synchronized resource cache
Posted by Piotr Kołaczkowski <pk...@gmail.com>.
Hi,
I don't think this would help. If a thread iterates over a collection
while another thread is modifying it, you'd get
ConcurrentModificationException anyway.
Is this iterator ever used somewhere? If it is, we have to solve it
somehow differently. Nevertheless making just puts and gets
synchronized solved the issue at least for us - we don't encounter any
crashes while starting the app.
The scenario that was causing this was the following:
Thread one: load window 1 from BXML
Thread two: load window 2 from BXML
Thread three: load window 3 from BXML
Thread four: load window 4 from BXML
Join threads
In GUI thread - open all the loaded windows
It is important for us to make it work (stable), because on a
multicore machine it improves loading windows by about 40% (not
perfect, but visibly faster).
BTW: Have you ever optimised pivot for faster startup times? The first
thing we noticed when we switched from Swing is that Pivot was somehow
slower to start up. This is worrying me - now it is not a problem it
the user waits for 1 second to load the app (I'm talking about warm
start up), but if in future the gui gets bigger... The competition
using JavaScript doesn't have that problem. :( Maybe there are some
possibilities to limit the amount of classloading or make some things
more lazy?
Thanks,
Piotr
2012/5/14 Sandro Martini <sa...@gmail.com>:
> Hi Piotr,
> if you want I can apply it ...
>
> but I think we need to synchronize even this method, right ?
>
> public synchronized Iterator<URL> iterator() {
> return new Iterator<URL>() {
> private Iterator<URI> iterator = resourceCache.iterator();
>
>
> Tell me ...
>
> Bye
Re: Synchronized resource cache
Posted by Piotr Kołaczkowski <pk...@gmail.com>.
Ok.
2012/5/13 Roger and Beth Whitcomb <Ro...@rbwhitcomb.com>:
> Hi Piotr,
> I don't see the patch, maybe make a JIRA issue and attach the patch there.
>
> ~Roger
>
> P.S. Welcome on board as a committer!
>
>
> On 5/13/12 6:50 AM, Piotr Kołaczkowski wrote:
>>
>> Hi,
>>
>> I noticed that sometimes when we load components from BXML in
>> parallel, we get weird errors in resource cache. Making it
>> synchronized seems to fix it.
>> Patch attached.
>>
>> Thanks,
>> Piotr
Re: Synchronized resource cache
Posted by Roger and Beth Whitcomb <Ro...@rbwhitcomb.com>.
Hi Piotr,
I don't see the patch, maybe make a JIRA issue and attach the patch there.
~Roger
P.S. Welcome on board as a committer!
On 5/13/12 6:50 AM, Piotr Kołaczkowski wrote:
> Hi,
>
> I noticed that sometimes when we load components from BXML in
> parallel, we get weird errors in resource cache. Making it
> synchronized seems to fix it.
> Patch attached.
>
> Thanks,
> Piotr