<html>
<head>
<base href="https://bugs.webkit.org/" />
</head>
<body>
<p>
<div>
<b><a class="bz_bug_link
bz_status_NEW "
title="NEW - Add acquire/release flags to B3::MemoryValue"
href="https://bugs.webkit.org/show_bug.cgi?id=162349#c3">Comment # 3</a>
on <a class="bz_bug_link
bz_status_NEW "
title="NEW - Add acquire/release flags to B3::MemoryValue"
href="https://bugs.webkit.org/show_bug.cgi?id=162349">bug 162349</a>
from <span class="vcard"><a class="email" href="mailto:fpizlo@apple.com" title="Filip Pizlo <fpizlo@apple.com>"> <span class="fn">Filip Pizlo</span></a>
</span></b>
<pre>I think it's important to review the optimizations we want to be able to do.
On x86, we want to be able to optimize store-load fences by transforming this:
store
fenceRW
load
Into this:
atomic store
load
We also want to turn Depend into a load-load fence, which reduces to no-op later.
On ARM we want to be able to optimize fences also. For example, we want to optimize load-load fences into load acquires:
load
fenceW
load
into:
load
load acquire
We want to be able to optimize store-load fences like this:
store
fenceRW
load
into:
store release
load acquire
Conversely, we want to be able to lower acq/rel flags. I believe that we want both store and load to accept such flags. That means that if we have a store acquire or load release, we want to be able to insert fences.</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are the assignee for the bug.</li>
</ul>
</body>
</html>