7077312: Provide a CALL effect for instruct declaration in the ad file
Summary: abstracted way to declare that the MachNode has the effect of a call (kills caller save registers, preserves callee save registers)
Reviewed-by: twisti, never
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN">
<html>
<head>
<title>Hotspot SA User Interface Notes</title>
</head>
<body>
<h1>Hotspot SA User Interface Notes</h1>
<h2>Workspace and Building</h2>
<p>
All the source code for the Serviceability Agent is in
<code>src/share/vm/agent</code> in the HotSport workspace
<code>/net/jano.sfbay/export/disk05/hotspot/ws/1.4/sa_baseline</code>
<p>
You can build the project by typing <code>gnumake</code> in the
<code>src/share/vm/agent</code> directory.
<p>
You can also use the default build target using the Ant build file (build.xml). You can download Ant from
<a href="http://jakarta.apache.org/ant">http://jakarta.apache.org/ant</a>. Documentation for Ant can be
found at <a href="http://jakarta.apache.org/ant/manual/index.html">http://jakarta.apache.org/ant/manual/index.html</a>
<h2>Running the project</h2>
<ul>
<li><code>java -cp classes sun.jvm.hotspot.HSDB</code>
<li><code>java -cp classes sun.jvm.hotspot.bugspot.Main</code>
</ul>
<h2>Feedback</h2>
<p>
Refactoring of package hierarchy. All user interface components should be in
the ui package. Perhaps: sun.jvm.hotspot.ui.hsdb.Main for the HSDB and
sun.jvm.hotspot.ui.bugspot.Main for BugSpot.
<p>
The src\share\vm\agent area seems like a workspace so it should be organized like
one. In particular, I'd like to suggest the following directory layout:<br>
<ul>
<li>src: All sources that are curently under the sun directory.
<li>classes: compiled class files.
<li>lib: Resources like images, icons and jar files.
<li>docs: Documentation
<li>deploy: distribution bundles for Java Web Start.
</ul>
<p>
Seems like there is a lot of redundant functionality. Between the HSDB and BugSpot. Perhaps
this can be consolidated with a <code>javax.swing.Actions</code> architecture.
<h2>Tasklist</h2>
<p>
<b>Stack memory pane</b>:
It's one of the more useful JVM debugging tools in the SA. However, it
doesn't support any interaction with the text; the Memory Panel in BugSpot
was written afterward (with help from Shannon) and implements proper
selection, scrolling, and drag-and-drop, but no annotations. I'm not sure how
to integrate the annotations with the JTable that's being used for the memory
view; if you have suggestions here please let me know.
<p>
<b>Integrations with the NetBeans architecture (plug in).</b> See the
<a href="http://openide.netbeans.org">Netbeans Open APIs homepage</a>
<p>
HSDB: Object Histogram. Column sizes should be sized according the the
contents. i.e, The size and count columns should be narrow enought to
handle the largest window. Since there is a lot of data, sorting
and searching should be implemented.
<p>
<h2>Log</h2>
<i>Last modified: Tue Feb 05 19:15:12 Pacific Standard Time 2002</i>
<p>
sun.jvm.hotspot.oops.ObjectHistogram should be the underlying data
structure for the TableModels. It shouldnt bother with sorting the data -
the table model should do that. It should implement these methods:
<pre>
public int getSize()
public ObjectHistogramElement getElementAt(int row);
</pre>
<p>
ObjectHistogramElement should return the String that represents
the third column
<hr>
<address><a href="mailto:mark.davidson@sun.com">Mark Davidson</a></address>
<!-- Created: Mon Jan 28 14:33:47 Pacific Standard Time 2002 -->
<!-- hhmts start -->
Last modified: Tue Feb 05 20:05:13 Pacific Standard Time 2002
<!-- hhmts end -->
</body>
</html>