On Thu, 18 Feb 2021 07:59:45 GMT, Andrey Turbanov 
<github.com+741251+turban...@openjdk.org> wrote:

>> Hello @turbanoff, the double read is necessary to correctly avoid any race 
>> conditions and is a typical strategy used in cases like these. I am not 
>> aware of any other alternate more performant strategy, for code like this.
>
> Let me be more clear: I think that it's enough to have only 2 `volatile` 
> field reads _in worst case_. We can use local variable to avoid more than 
> needed reads.
> Current code can perform read 4 times _in worst case_: twice outside 
> `synchronized` block and twice inside `synchronized` block.
> 
> There are many examples of similar code in the JDK:
> https://github.com/openjdk/jdk/blob/master/src/jdk.crypto.cryptoki/share/classes/sun/security/pkcs11/P11Util.java#L48
> https://github.com/openjdk/jdk/blob/master/src/java.base/share/classes/java/lang/StackFrameInfo.java#L125

Ah! of course, I now see what you mean. I misunderstood your previous comment. 
I'll update this PR shortly.

-------------

PR: https://git.openjdk.java.net/jdk/pull/2601

Reply via email to