RFR (XS): 8036860: Pad and cache-align the BiasedMappedArray

[Stefan Karlsson]

On 2014-03-14 10:35, Thomas Schatzl wrote:
> Hi,
>
>    I messed up the webrev link (and the CR number in the topic) due to
> two very similarly named CRs I currently have assigned to me.
>
> Here is the correct webrev URL:
>
> http://cr.openjdk.java.net/~tschatzl/8036860/webrev/

http://cr.openjdk.java.net/~tschatzl/8036860/webrev/src/share/vm/gc_implementation/g1/g1BiasedArray.hpp.udiff.html

Could you move create_new_base_array(size_t length, size_t elem_size) to 
the .cpp file, so that we don't have to include padded.inline.hpp in the 
.hpp file?

The allocation.inline.hpp include should probably be chanted to 
allocation.hpp.


http://cr.openjdk.java.net/~tschatzl/8036860/webrev/src/share/vm/memory/padded.inline.hpp.udiff.html

+  return (T*)align_size_up_((uintptr_t)chunk, alignment);

Could you use align_pointer_up instead, as you did in 8035815?


Other than that, this looks good.

thanks,
StefanK

>
> On Fri, 2014-03-14 at 10:25 +0100, Thomas Schatzl wrote:
>> Hi all,
>>
>>    can I have reviews for the following tiny change? It adds padding for
>> all BiasedMappedArray instances.
>>
>> Instances of those are frequently used in hot loops (or are going to,
>> see JDK-8027553), so it seems a good idea to let them start on cache
>> line boundaries and pad the tail.
>>
>> Webrev:
>> http://cr.openjdk.java.net/~tschatzl/8035815/webrev/
>>
>> CR:
>> https://bugs.openjdk.java.net/browse/JDK-8036860
>>
>> Testing:
>> jprt
> Sorry,
>    Thomas
>