Can virtual thread get better support for Map/Reduce like thing?

[何品(虎鸣)]

The current implementation always adds the task to the tail, but what if a map/reduce like work, which mean a fork() maybe better for performance/latency , but that will reduce some fairness.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/loom-dev/attachments/20240621/1f00bbc0/attachment.htm>