[External] : Re: Adaptation of IO classes to the VTs

[Andrii Lomakin]

|  Yes. As a matter of fact, we’re exploring two solutions simultaneously.
One with less-than-ideal performance, which could hopefully be delivered
sooner, and another with ideal performance but that will likely take longer
to deliver. It’s possible we’ll deliver the short term solution first, and
the long-term solution later.
Cool, hope to read more details about those implementations once they are
ready!

On Wed, Jul 5, 2023 at 2:04 PM Ron Pressler <ron.pressler at oracle.com> wrote:

>
>
> > On 5 Jul 2023, at 12:58, Andrii Lomakin <lomakin.andrey at gmail.com>
> wrote:
> >
> > Hi Ron.
> > Thank you for getting back to me so quickly.
> >
> > Yep, it slipped my mind that ByteInputStream calls are short and not
> blocking, the first example I remembered.
> >
> > Follow-up questions:
> > 1. By tracking commits in the loom branch, I have had the impression
> that working on the ability not to pin threads in synchronized blocks has
> already started. Is this the correct impression?
>
> Yes. As a matter of fact, we’re exploring two solutions simultaneously.
> One with less-than-ideal performance, which could hopefully be delivered
> sooner, and another with ideal performance but that will likely take longer
> to deliver. It’s possible we’ll deliver the short term solution first, and
> the long-term solution later.
>
> > 2. "Yes. The work is ongoing." - Just to double check, is my
> understanding correct that once this work is done, the worker thread (OS
> thread) will not be frozen by IO calls? I am interested in calls done by
> FileChannel,  etc.
>
> Yes, where io_uring is available.
>
> — Ron



-- 
Best regards,
Andrii Lomakin.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/loom-dev/attachments/20230705/224271a7/attachment.htm>