<!DOCTYPE html><html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<br>
<br>
<div class="moz-cite-prefix">On 12/07/2024 17:46, Konstantin Nisht
wrote:<br>
</div>
<blockquote type="cite" cite="mid:CAMQu2QBhh_Y0YHjVWhmx+7hnM70K3EUiNy+=U9ahYPnmUtSB1g@mail.gmail.com">
<div dir="ltr">
<div>
<div>> Is there a brief summary on what the performance is?<br>
</div>
It takes 1 minute 20 seconds to open a 12 kb zip file located
on the Linux side of WSL from Windows. The same file can be
opened in 0.02 seconds if located on Windows.<br>
This problem is in the implementation of 9P by Microsoft, but
it is quite difficult to workaround it, especially when we
have clients that use pure <a href="https://urldefense.com/v3/__http://java.io__;!!ACWV5N9M2RV99hQ!KowmBF66mgxvmW_n3a0ATp5w0P9Go4mGHYtZF_iLFsfLWCc9J3qA4t8f_BiQ-1OBthJeiIMQ4dNkEG9w11sY0eQ7gcRsqZg$" moz-do-not-send="true">java.io</a>. <br>
<br>
<span class="gmail-im">> </span>The motive for these
prototypes is virtual threads.<br>
</div>
Could you clarify how FileChannel helps here?<br>
</div>
</blockquote>
<br>
Essentially the same reason that Socket, DatagramSocket etc. were
changed to use NIO with JEP 353 JEP 373. There is significant
surgery requirement to allow virtual threads relase their carrier
for file operations. This is not something for the short term of
course, I'm just pointing out other efforts to re-implement RAF that
is not observable except a very small startup regression at this
point.<br>
<br>
-Alan<br>
</body>
</html>