<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<br>
<div class="moz-cite-prefix">On 4/29/2015 4:31 PM, Jungwoo Ha wrote:<br>
</div>
<blockquote
cite="mid:CA+n_jhiSP9ELRrg9fC_Ci9_WVS02DK6EkJ4HdreaGkzF5a9OFw@mail.gmail.com"
type="cite">
<div dir="ltr">Alright. I'll send it to runtime. <br>
</div>
</blockquote>
<br>
Jungwoo,<br>
<br>
Let me suggest hotspot-dev. But if you've already sent it,<br>
runtime is OK too.<br>
<br>
Jon<br>
<br>
<blockquote
cite="mid:CA+n_jhiSP9ELRrg9fC_Ci9_WVS02DK6EkJ4HdreaGkzF5a9OFw@mail.gmail.com"
type="cite">
<div class="gmail_extra"><br>
<div class="gmail_quote">On Wed, Apr 29, 2015 at 4:27 PM, Kim
Barrett <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:kim.barrett@oracle.com" target="_blank">kim.barrett@oracle.com</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="HOEnZb">
<div class="h5">On Apr 29, 2015, at 5:46 PM, Jungwoo Ha
<<a moz-do-not-send="true"
href="mailto:jwha@google.com">jwha@google.com</a>>
wrote:<br>
><br>
> Oops. I meant to copy BUG-8079091<br>
> <a moz-do-not-send="true"
href="http://cr.openjdk.java.net/%7Ejwha/8079091/webrev.00/"
target="_blank">http://cr.openjdk.java.net/~jwha/8079091/webrev.00/</a><br>
><br>
><br>
> On Wed, Apr 29, 2015 at 2:40 PM, Jungwoo Ha <<a
moz-do-not-send="true" href="mailto:jwha@google.com">jwha@google.com</a>>
wrote:<br>
> BUG: <a moz-do-not-send="true"
href="https://bugs.openjdk.java.net/browse/JDK-8079091"
target="_blank">https://bugs.openjdk.java.net/browse/JDK-8079091</a><br>
> Webrev: <a moz-do-not-send="true"
href="http://cr.openjdk.java.net/%7Ejwha/8075288/webrev.00/"
target="_blank">http://cr.openjdk.java.net/~jwha/8075288/webrev.00/</a><br>
><br>
> Can someone sponsor this change?<br>
> I've seen ~3% speed up on DaCapo benchmarks.
(results attached on the bug page.)<br>
> Regardless of the speed up, I think the changes are
straight-forward improvement.<br>
><br>
> --<br>
> Jungwoo Ha<br>
<br>
</div>
</div>
I'm not sure of this, but I think metaspace belongs to
runtime rather<br>
than gc. If so, this should go through the hs-rt repository
and have<br>
runtime folks involved in the review.<br>
<span class=""><br>
> I've seen ~3% speed up on DaCapo benchmarks. (results
attached on the bug page.)<br>
<br>
</span>Wow!<br>
<br>
Hm. There are a couple of benchmarks that are a little bit
slower. I<br>
don't know enough about that suite of benchmarks to know how
important<br>
that might be.<br>
<br>
Since the dictionary is always being created at construction
time, the<br>
null check in ~BlockFreelist is no longer needed either.<br>
<br>
It seems like a further improvement might be obtained by
making the<br>
dictionary a direct member subobject of BlockFreelist,
rather than<br>
allocating it as a separate object, thereby eliminating an<br>
indirection.<br>
<br>
Pre-existing: Why is there a cast in
SpaceManager::block_freelists()?<br>
<br>
Pre-existing: I found the private
BlockFreelist::dictionary() member<br>
function distracting rather than helpful, and would have
preferred<br>
just using the private data member directly. But maybe
that's just me.<br>
If _dictionary is changed to a subobject rather than a
pointer, I'd<br>
suggest changing a bunch of "dictionary()->XXX" to
"_dictionary.XXX".<br>
<br>
</blockquote>
</div>
<br>
</div>
</blockquote>
<br>
</body>
</html>