Author: bob

[This entry backdated to when I actually wrote it]

I’m currently sitting in the Atlanta airport as part of a quick in-and-out to visit a sick relative. I’d easily give a digit for a sausage biscuit right now, but alas the best I can do is trail mix.

More evidence that the terrorists have won: while waiting in the security line with 2000 of my closest friends, having disposed of the Purell I forgot to check, they started playing some Glenn Medeiros over the intercom. You know,

Nothing’s gonna change my love for you
You ought know by now how much I love you
One thing you can be sure of
I’ll never ask for more than your love

There, now you have it stuck in your head too.

I’m reading Kitchen Confidential on the strength of this blog entry by “No Reservations” star Anthony Bourdain. It’s a humorous look at how the restaurant industry is much less glamorous than one might think. [Later on the plane, my single-serving friend, a philosophical sheet metal worker, said this is like life: first you want to get with some beautiful woman, then you realize she’s nasty inside, but by then it’s too late because you’re already married.]

In other news, Angeline and I saw a famous celebrity while dining at the White Flint mall the other night: Matthew Lesko, the question-mark-blazer-ensuited guy who exhorts us to get FREE government money. He was in uniform and looked lost. Doesn’t ring a bell? Here’s Andy Dick’s impression of him.

Congratulations, NASA, on finally managing to make space interesting again!

In other last week’s news, bunnie of Xbox hackery fame has a nice rundown on why the ATHF debacle was such a sad commentary on our society. I hope the 2 mil extortion that Ted agreed to is just a face-saving thing and no money actually changed hands.

I can’t remember what I was going to write here, except that I suck at Monopoly. We had a good time in Toronto even though it was crowded with eight people in a 2BR apartment. The weather was nice and warm. Weird, eh? Speaking of “eh,” I only heard one person say it, while we were watching some hockey game in a Queen Street pool hall. I came back with only a few CDN dollars this time. They don’t go quite as far as they used to.

Here’s Christmas 2006, in thumbnails. The rest of my pictures are somewhere in Baltimore.

Angeline & I went up to the not-quite-frozen north last week to visit her fam for Christmas. I shall account for time spent there in a future entry a few days from now, but not right now because the pictures I want to post are on my camera that I left there.

Instead, I’ll continue to post about boring technical subjects. Today: one example of sucky Java performance (a topic which hardly anyone in the Java community cares about, apparently). I had a disagreement of opinion with a co-worker recently, as I was coming across code written by a 3rd person that went like:

public void getBar() { return this.bar; }
public void doSomething() {
getBar().frob();
getBar().xyzzy();
getBar().baz();
}

I remarked that this sort of programming is not only extra typing, but performs poorly because inside doSomething we are now making six method calls instead of only the three we need. My co-worker suggested the Java compiler was smart enough to inline getBar(). I was skeptical for two reasons: first that getBar is public, so in order to allow subclasses to override it, it has to be in the Java-equivalent of a vtable, like every other Java method. Second, I’ve looked at the output of javac and it is really quite dumb.

So, I decided to make it easy on the compiler in a test program to see (excerpt):

private static final String blah="blah";
private final String getBlah() { return blah; }
public void doit() { System.out.println(getBlah()); }

There’s no way for a subclass to override getBlah or the value it returns; however, the code generated is:

public void doit();
Code:
0:	getstatic	#3; //Field java/lang/System.out:Ljava/io/PrintStream;
3:	aload_0
4:	invokespecial	#4; //Method getBlah:()Ljava/lang/String;
7:	invokevirtual	#5; //Method java/io/PrintStream.println:(Ljava/lang/String;)V
10:	return

Obviously, the final keyword does nothing. Now, I don’t know much about how the JVM/JIT will further optimize this, but this can’t be more efficient than a simple load of the variable onto the stack. We have to store this on the stack for one thing, and the JVM needs to store a return address somewhere. A small effect, sure, but why throw away performance?

Maybe I’ll be evil and start asking why Java should have the ‘virtual’ keyword in interviews.