Re: RFR: 8171373: Reduce copying during initialization of ModuleHashes

[Mandy Chung]

+1

Mandy

> On Dec 16, 2016, at 4:45 PM, Claes Redestad  wrote:
> 
> Chris, Mandy,
> 
> thanks for reviewing!
> 
> I've made nameToHash final and added a comment, updated in place.
> 
> I'll push this tomorrow unless I hear any objections.
> 
> /Claes
> 
> On 2016-12-16 18:52, Mandy Chung wrote:
>> 
>>> On Dec 16, 2016, at 8:00 AM, Claes Redestad  
>>> wrote:
>>> 
>>> Hi,
>>> 
>>> recent changes to split out ModuleHashes from ModuleDescriptor caused a 
>>> small/tiny increase in HashMap creation, resize and copying:
>>> 
>>> Webrev: http://cr.openjdk.java.net/~redestad/8171373/webrev.01/
>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8171373
>> 
>> I like the optimization for one single module (i.e. java.base) containing 
>> the module hashes which is the common case.
>> 
>> Can you add a comment in SystemModuleFinder about the common case and avoid 
>> creating a new HashMap unless there is more than one module containing 
>> ModuleHashes.
>> 
>> Builder::nameToHash can now be final.
>> 
>> Mandy
>> 

Re: RFR: 8171373: Reduce copying during initialization of ModuleHashes

[Claes Redestad]

Chris, Mandy,

thanks for reviewing!

I've made nameToHash final and added a comment, updated in place.

I'll push this tomorrow unless I hear any objections.

/Claes

On 2016-12-16 18:52, Mandy Chung wrote:



On Dec 16, 2016, at 8:00 AM, Claes Redestad  wrote:

Hi,

recent changes to split out ModuleHashes from ModuleDescriptor caused a 
small/tiny increase in HashMap creation, resize and copying:

Webrev: http://cr.openjdk.java.net/~redestad/8171373/webrev.01/
Bug: https://bugs.openjdk.java.net/browse/JDK-8171373


I like the optimization for one single module (i.e. java.base) containing the 
module hashes which is the common case.

Can you add a comment in SystemModuleFinder about the common case and avoid 
creating a new HashMap unless there is more than one module containing 
ModuleHashes.

Builder::nameToHash can now be final.

Mandy

Re: RFR: 8171373: Reduce copying during initialization of ModuleHashes

[Mandy Chung]

> On Dec 16, 2016, at 8:00 AM, Claes Redestad  wrote:
> 
> Hi,
> 
> recent changes to split out ModuleHashes from ModuleDescriptor caused a 
> small/tiny increase in HashMap creation, resize and copying:
> 
> Webrev: http://cr.openjdk.java.net/~redestad/8171373/webrev.01/
> Bug: https://bugs.openjdk.java.net/browse/JDK-8171373

I like the optimization for one single module (i.e. java.base) containing the 
module hashes which is the common case.

Can you add a comment in SystemModuleFinder about the common case and avoid 
creating a new HashMap unless there is more than one module containing 
ModuleHashes.

Builder::nameToHash can now be final.

Mandy

Re: RFR: 8171373: Reduce copying during initialization of ModuleHashes

[Chris Hegarty]

> On 16 Dec 2016, at 16:00, Claes Redestad  wrote:
> 
> Hi,
> 
> recent changes to split out ModuleHashes from ModuleDescriptor caused a 
> small/tiny increase in HashMap creation, resize and copying:
> 
> Webrev: http://cr.openjdk.java.net/~redestad/8171373/webrev.01/
> Bug: https://bugs.openjdk.java.net/browse/JDK-8171373
> 
> This decreases #bytecode executed during startup by ~20K (around 1.5% of 
> total).

Nicely done Claes. Looks good to me.

Builder.nameToHash can be final, no?

-Chris.

RFR: 8171373: Reduce copying during initialization of ModuleHashes

[Claes Redestad]

Hi,

recent changes to split out ModuleHashes from ModuleDescriptor caused a 
small/tiny increase in HashMap creation, resize and copying:


Webrev: http://cr.openjdk.java.net/~redestad/8171373/webrev.01/
Bug: https://bugs.openjdk.java.net/browse/JDK-8171373

This decreases #bytecode executed during startup by ~20K (around 1.5% of 
total).


Thanks!

/Claes