Hi,
Thanks for infos.

By the way I noticed netty had a variation of this class in their code
which inits seed differently.

Regards



On Mon, Jan 26, 2015 at 7:28 AM, Felix Schumacher <
[email protected]> wrote:

>
>
> Hi Phillipe.
>
> Am 25. Januar 2015 22:42:27 MEZ, schrieb Philippe Mouawad <
> [email protected]>:
> >Hi Felix,
> >Reverted , can you explain to me this ? I don't see these fields used,
> >are
> >they through reflection ?
>
> They are not used. Neither directly nor indirectly. They are used as
> padding to prevent cache issues on hardware level. You will find more
> information when you search the Internet for "ThreadLocalRandom padding".
>
> It might be, that we don't need the padding on modern jvms, but if we
> assumed to have those, we would not need the copy of the class either :)
>
> That is why I opt for keeping the "useless" variables.
>
> Regards
> Felix
> >
> >Thanks
> >
> >On Sun, Jan 25, 2015 at 10:30 PM, Felix Schumacher <
> >[email protected]> wrote:
> >
> >>
> >>
> >> Am 25. Januar 2015 21:35:48 MEZ, schrieb [email protected]:
> >> >Author: pmouawad
> >> >Date: Sun Jan 25 20:35:48 2015
> >> >New Revision: 1654701
> >> >
> >> >URL: http://svn.apache.org/r1654701
> >> >Log:
> >> >Remove unused fields
> >> >
> >> >Modified:
> >> >
> >jmeter/trunk/src/core/org/apache/jmeter/util/ThreadLocalRandom.java
> >> >
> >> >Modified:
> >> >jmeter/trunk/src/core/org/apache/jmeter/util/ThreadLocalRandom.java
> >> >URL:
> >> >
> >>
> >
> http://svn.apache.org/viewvc/jmeter/trunk/src/core/org/apache/jmeter/util/ThreadLocalRandom.java?rev=1654701&r1=1654700&r2=1654701&view=diff
> >>
> >>
>
> >>==============================================================================
> >> >---
> >jmeter/trunk/src/core/org/apache/jmeter/util/ThreadLocalRandom.java
> >> >(original)
> >> >+++
> >jmeter/trunk/src/core/org/apache/jmeter/util/ThreadLocalRandom.java
> >> >Sun Jan 25 20:35:48 2015
> >> >@@ -54,11 +54,6 @@ public class ThreadLocalRandom extends R
> >> >      */
> >> >     boolean initialized;
> >> >
> >> >-    // Padding to help avoid memory contention among seed updates
> >in
> >> >-    // different TLRs in the common case that they are located near
> >> >-    // each other.
> >> >-    private long pad0, pad1, pad2, pad3, pad4, pad5, pad6, pad7;
> >>
> >> Given the comment above the fields, I would not remove those padding
> >> fields.
> >>
> >> Regards
> >> Felix
> >> >-
> >> >     /**
> >> >      * The actual ThreadLocal
> >> >      */
> >>
> >>
>
>


-- 
Cordialement.
Philippe Mouawad.

Reply via email to