Subject: Re: C435 JGFSI Reviews
Resent-Date: Fri, 12 Nov 1999 08:37:58 -0500
Resent-From: Geoffrey Fox <gcf@npac.syr.edu>
Resent-To: p_gcf@npac.syr.edu
Date: Fri, 5 Nov 1999 23:47:31 -0600
From: Joel Jones <jjones@uiuc.edu>
To: Geoffrey Fox <gcf@npac.syr.edu>

   I've gone over the reviewer's comments and here is what I've done
with regard to the rewrite.  I have enclosed postscript and pdf
formats.  I did this in LaTeX, so if you want that instead, just let
me know.

Reviewer 1:
   He liked my work, so there's not much to do here.  :)
   o changed the wording in my description of the IBM Tokyo work with regard to
     doing graph-coloring register allocation in the VM.
   o Add a better description of why the swap is placed on the bytecode at PC 30

Reviewer 3:
   His comments about writing style I have taken with a grain of salt,
as it directly contradicts reviewer 1.  However, there are a few
points rising from his comments that I have addressed.

   o I don't think a more detailed description of my register allocation
     algorithm is appropriate here.  My thinking is that anyone who is
     familiar with the Brigg's-Chaitin approach will probably be able to
     understand what my modifications are.  If they aren't, then I really don't
     want to reproduce the ~15 pages of description that Briggs has in this
     thesis.

   o I made a comment emphasizing that adding the VR annotation
     doesn't preclude the use of other bytecode transformations done by the JVM.
     It merely changes the IR from bytecode for a stack machine to bytecode for
     an infinite register machine.

   o I mention the names and purposes of the non-VR annotations sooner
     in the paper.  I think this will help clear up some of the confusion he had
     with section 4.

   o Changed the section 3 title from "Motivating Example" to
     "Simple Example", to indicate that the example is not meant to show the
     benefit, so much as the environment.  I also added text along
those lines at
     the beginning of the section.

Reviewer 2:
   The second reviewer's comments were more substantive and to the
point.  We really haven't answered many of the important questions.
None of these are questions that we hadn't already known---I just
didn't raise them since I have an immature implementation at the
moment.  I dealt with this issue by adding an "Open Issues" section
to discuss these questions.  I think that this is a valid way to deal
with the reviewer's concerns because there exist other researchers
whose work also hasn't answered these questions yet.  This helps to
lay rest to the "they haven't thought about the questions raised by
their work and nobody cares about the questions anyway" sort of
comments.

I specifically did not deal with his comments on inlining.  I would
prefer not to offer up my ideas on inlining and annotations at this
time.  First, and most important, they are still very immature.
Second, I don't want to let my ideas out too long before I have an
opportunity to exploit them.

Joel Jones
jjones@uiuc.edu

                  Name: CPE1999.pdf
   CPE1999.pdf    Type: Acrobat (application/pdf)
              Encoding: base64

                 Name: CPE1999.ps
   CPE1999.ps    Type: Postscript Document (application/postscript)
             Encoding: base64