(L) Prelim RFR: 8132510: Replace ThreadLocalStorage with compiler/language-based thread-local variables

Andrew Haley aph at redhat.com
Fri Nov 6 10:57:30 UTC 2015


On 06/11/15 10:09, David Holmes wrote:
> On 6/11/2015 7:53 PM, Andrew Haley wrote:
>> On 05/11/15 22:54, David Holmes wrote:
>>> Not sure why you had to inject the helper rather than just "call"
>>> Thread::current()
>>
>> Because Thread::current() might clobber vector state.  It's perfectly
>> well allowed to.
> 
> Sorry I really don't understand. If Thread::current can clobber state 
> then your helper will have to preserve it. So why can't the caller of 
> the helper do the same preservation and skip the need for a helper??

My helper is written in assembly language.  It is not written in C.
It does not call Thread::current().  It accesses the _thread variable
directly.  It does not clobber vector state.

Andrew.



More information about the hotspot-dev mailing list