Re: [abc] ecoop paper

On Fri, 17 Dec 2004, Sascha Kuzins wrote:

> I just went over the new section. Some points:
> do we need pure java numbers for the pure java benchmarks?

with javac, you mean? Yes, that would be good.

> abc -O0 is slower than abc, I'm about to fix the script.

ok.

> readers may wonder if abc programs are faster than ajc+soot programs.

the answer is yes, of course. I'll mention it in the runtime section.

> We don't mention the size of soot (but we do mention BCEL).

That is because BCEL is part of the ajc code base, but Soot is not
part of abc.

> 6.3 a stronger distinction between slots and local variables would be nice

> 6.4 10000% => 100x?

Yes, that's right (figure).

-O
Received on Fri Dec 17 12:10:48 2004